fix rounding bug

* fix update preview bug

* fix update preview bug

* remove guard

* add is_vision_frame

.coveragerc

@@ -0,0 +1,2 @@
+[run]
+patch = subprocess

.github/workflows/ci.yml

@@ -33,7 +33,7 @@ jobs:
     uses: actions/setup-python@v5
     with:
      python-version: '3.12'
-  - run: python install.py --onnxruntime default --skip-conda
+  - run: python install.py default --skip-conda
   - run: pip install pytest
   - run: pytest
  report:
@@ -48,7 +48,7 @@ jobs:
     uses: actions/setup-python@v5
     with:
      python-version: '3.12'
-  - run: python install.py --onnxruntime default --skip-conda
+  - run: python install.py default --skip-conda
   - run: pip install coveralls
   - run: pip install pytest
   - run: pip install pytest-cov

.gitignore

@@ -1,6 +1,7 @@
 __pycache__
 .assets
+.claude
 .caches
-.jobs
 .idea
+.jobs
 .vscode

LICENSE.md

@@ -1,3 +1,3 @@
 OpenRAIL-AS license
 
-Copyright (c) 2025 Henry Ruhs
+Copyright (c) 2026 Henry Ruhs

README.md

@@ -37,6 +37,7 @@ commands:
     headless-run                                  run the program in headless mode
     batch-run                                     run the program in batch mode
     force-download                                force automate downloads and exit
+    benchmark                                     benchmark the program
     job-list                                      list jobs by status
     job-create                                    create a drafted job
     job-submit                                    submit a drafted job to become a queued job

facefusion.ini

@@ -1,3 +1,7 @@
+[workflow]
+workflow_mode =
+workflow_strategy =
+
 [paths]
 temp_path =
 jobs_path =
@@ -13,6 +17,7 @@ output_pattern =
 [face_detector]
 face_detector_model =
 face_detector_size =
+face_detector_margin =
 face_detector_angles =
 face_detector_score =
 
@@ -31,40 +36,54 @@ reference_face_position =
 reference_face_distance =
 reference_frame_number =
 
+[face_tracker]
+face_tracker_score =
+
 [face_masker]
 face_occluder_model =
 face_parser_model =
 face_mask_types =
+face_mask_areas =
+face_mask_regions =
 face_mask_blur =
 face_mask_padding =
-face_mask_regions =
+
+[voice_extractor]
+voice_extractor_model =
 
 [frame_extraction]
 trim_frame_start =
 trim_frame_end =
 temp_frame_format =
-keep_temp =
+temp_pixel_format =
+
+[frame_distribution]
+target_frame_amount =
 
 [output_creation]
 output_image_quality =
-output_image_resolution =
+output_image_scale =
 output_audio_encoder =
 output_audio_quality =
 output_audio_volume =
 output_video_encoder =
 output_video_preset =
 output_video_quality =
-output_video_resolution =
+output_video_scale =
 output_video_fps =
 
 [processors]
 processors =
 age_modifier_model =
 age_modifier_direction =
+background_remover_model =
+background_remover_fill_color =
+background_remover_despill_color =
 deep_swapper_model =
 deep_swapper_morph =
 expression_restorer_model =
 expression_restorer_factor =
+expression_restorer_areas =
 face_debugger_items =
 face_editor_model =
 face_editor_eyebrow_direction =
@@ -86,31 +105,36 @@ face_enhancer_blend =
 face_enhancer_weight =
 face_swapper_model =
 face_swapper_pixel_boost =
+face_swapper_weight =
 frame_colorizer_model =
 frame_colorizer_size =
 frame_colorizer_blend =
 frame_enhancer_model =
 frame_enhancer_blend =
 lip_syncer_model =
+lip_syncer_weight =
 
 [uis]
 open_browser =
 ui_layouts =
 ui_workflow =
 
-[execution]
-execution_device_id =
-execution_providers =
-execution_thread_count =
-execution_queue_count =
-
 [download]
 download_providers =
 download_scope =
 
+[benchmark]
+benchmark_mode =
+benchmark_resolutions =
+benchmark_cycle_count =
+
+[execution]
+execution_device_ids =
+execution_providers =
+execution_thread_count =
+
 [memory]
 video_memory_strategy =
-system_memory_limit =
 
 [misc]
 log_level =

facefusion.py

@@ -4,7 +4,8 @@ import os
 
 os.environ['OMP_NUM_THREADS'] = '1'
 
-from facefusion import core
+from facefusion import conda, core
 
 if __name__ == '__main__':
+	conda.setup()
 	core.cli()

facefusion/app_context.py

@@ -5,12 +5,14 @@ from facefusion.types import AppContext
 
 
 def detect_app_context() -> AppContext:
+	jobs_path = os.path.join('facefusion', 'jobs')
+	uis_path = os.path.join('facefusion', 'uis')
 	frame = sys._getframe(1)
 
 	while frame:
-		if os.path.join('facefusion', 'jobs') in frame.f_code.co_filename:
+		if jobs_path in frame.f_code.co_filename:
 			return 'cli'
-		if os.path.join('facefusion', 'uis') in frame.f_code.co_filename:
+		if uis_path in frame.f_code.co_filename:
 			return 'ui'
 		frame = frame.f_back
 	return 'cli'

facefusion/args.py

@@ -1,10 +1,91 @@
 from facefusion import state_manager
-from facefusion.filesystem import get_file_name, is_image, is_video, resolve_file_paths
+from facefusion.filesystem import get_file_name, is_video, resolve_file_paths
 from facefusion.jobs import job_store
-from facefusion.normalizer import normalize_fps, normalize_padding
+from facefusion.normalizer import normalize_fps, normalize_space
 from facefusion.processors.core import get_processors_modules
 from facefusion.types import ApplyStateItem, Args
-from facefusion.vision import create_image_resolutions, create_video_resolutions, detect_image_resolution, detect_video_fps, detect_video_resolution, pack_resolution
+from facefusion.vision import detect_video_fps
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('command', args.get('command'))
+	apply_state_item('workflow_mode', args.get('workflow_mode'))
+	apply_state_item('workflow_strategy', args.get('workflow_strategy'))
+	apply_state_item('temp_path', args.get('temp_path'))
+	apply_state_item('jobs_path', args.get('jobs_path'))
+	apply_state_item('source_paths', args.get('source_paths'))
+	apply_state_item('target_path', args.get('target_path'))
+	apply_state_item('output_path', args.get('output_path'))
+	apply_state_item('source_pattern', args.get('source_pattern'))
+	apply_state_item('target_pattern', args.get('target_pattern'))
+	apply_state_item('output_pattern', args.get('output_pattern'))
+	apply_state_item('face_detector_model', args.get('face_detector_model'))
+	apply_state_item('face_detector_size', args.get('face_detector_size'))
+	apply_state_item('face_detector_margin', normalize_space(args.get('face_detector_margin')))
+	apply_state_item('face_detector_angles', args.get('face_detector_angles'))
+	apply_state_item('face_detector_score', args.get('face_detector_score'))
+	apply_state_item('face_landmarker_model', args.get('face_landmarker_model'))
+	apply_state_item('face_landmarker_score', args.get('face_landmarker_score'))
+	apply_state_item('face_selector_mode', args.get('face_selector_mode'))
+	apply_state_item('face_selector_order', args.get('face_selector_order'))
+	apply_state_item('face_selector_age_start', args.get('face_selector_age_start'))
+	apply_state_item('face_selector_age_end', args.get('face_selector_age_end'))
+	apply_state_item('face_selector_gender', args.get('face_selector_gender'))
+	apply_state_item('face_selector_race', args.get('face_selector_race'))
+	apply_state_item('reference_face_position', args.get('reference_face_position'))
+	apply_state_item('reference_face_distance', args.get('reference_face_distance'))
+	apply_state_item('reference_frame_number', args.get('reference_frame_number'))
+	apply_state_item('face_tracker_score', args.get('face_tracker_score'))
+	apply_state_item('face_occluder_model', args.get('face_occluder_model'))
+	apply_state_item('face_parser_model', args.get('face_parser_model'))
+	apply_state_item('face_mask_types', args.get('face_mask_types'))
+	apply_state_item('face_mask_areas', args.get('face_mask_areas'))
+	apply_state_item('face_mask_regions', args.get('face_mask_regions'))
+	apply_state_item('face_mask_blur', args.get('face_mask_blur'))
+	apply_state_item('face_mask_padding', normalize_space(args.get('face_mask_padding')))
+	apply_state_item('voice_extractor_model', args.get('voice_extractor_model'))
+	apply_state_item('trim_frame_start', args.get('trim_frame_start'))
+	apply_state_item('trim_frame_end', args.get('trim_frame_end'))
+	apply_state_item('temp_frame_format', args.get('temp_frame_format'))
+	apply_state_item('temp_pixel_format', args.get('temp_pixel_format'))
+	apply_state_item('target_frame_amount', args.get('target_frame_amount'))
+	apply_state_item('output_image_quality', args.get('output_image_quality'))
+	apply_state_item('output_image_scale', args.get('output_image_scale'))
+	apply_state_item('output_audio_encoder', args.get('output_audio_encoder'))
+	apply_state_item('output_audio_quality', args.get('output_audio_quality'))
+	apply_state_item('output_audio_volume', args.get('output_audio_volume'))
+	apply_state_item('output_video_encoder', args.get('output_video_encoder'))
+	apply_state_item('output_video_preset', args.get('output_video_preset'))
+	apply_state_item('output_video_quality', args.get('output_video_quality'))
+	apply_state_item('output_video_scale', args.get('output_video_scale'))
+
+	if args.get('output_video_fps') or is_video(args.get('target_path')):
+		output_video_fps = normalize_fps(args.get('output_video_fps')) or detect_video_fps(args.get('target_path'))
+		apply_state_item('output_video_fps', output_video_fps)
+
+	available_processors = [ get_file_name(file_path) for file_path in resolve_file_paths('facefusion/processors/modules') ]
+	apply_state_item('processors', args.get('processors'))
+
+	for processor_module in get_processors_modules(available_processors):
+		processor_module.apply_args(args, apply_state_item)
+
+	apply_state_item('open_browser', args.get('open_browser'))
+	apply_state_item('ui_layouts', args.get('ui_layouts'))
+	apply_state_item('ui_workflow', args.get('ui_workflow'))
+	apply_state_item('execution_device_ids', args.get('execution_device_ids'))
+	apply_state_item('execution_providers', args.get('execution_providers'))
+	apply_state_item('execution_thread_count', args.get('execution_thread_count'))
+	apply_state_item('download_providers', args.get('download_providers'))
+	apply_state_item('download_scope', args.get('download_scope'))
+	apply_state_item('benchmark_mode', args.get('benchmark_mode'))
+	apply_state_item('benchmark_resolutions', args.get('benchmark_resolutions'))
+	apply_state_item('benchmark_cycle_count', args.get('benchmark_cycle_count'))
+	apply_state_item('video_memory_strategy', args.get('video_memory_strategy'))
+	apply_state_item('log_level', args.get('log_level'))
+	apply_state_item('halt_on_error', args.get('halt_on_error'))
+	apply_state_item('job_id', args.get('job_id'))
+	apply_state_item('job_status', args.get('job_status'))
+	apply_state_item('step_index', args.get('step_index'))
 
 
 def reduce_step_args(args : Args) -> Args:
@@ -37,100 +118,3 @@ def collect_job_args() -> Args:
 		key: state_manager.get_item(key) for key in job_store.get_job_keys() #type:ignore[arg-type]
 	}
 	return job_args
-
-
-def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
-	# general
-	apply_state_item('command', args.get('command'))
-	# paths
-	apply_state_item('temp_path', args.get('temp_path'))
-	apply_state_item('jobs_path', args.get('jobs_path'))
-	apply_state_item('source_paths', args.get('source_paths'))
-	apply_state_item('target_path', args.get('target_path'))
-	apply_state_item('output_path', args.get('output_path'))
-	# patterns
-	apply_state_item('source_pattern', args.get('source_pattern'))
-	apply_state_item('target_pattern', args.get('target_pattern'))
-	apply_state_item('output_pattern', args.get('output_pattern'))
-	# face detector
-	apply_state_item('face_detector_model', args.get('face_detector_model'))
-	apply_state_item('face_detector_size', args.get('face_detector_size'))
-	apply_state_item('face_detector_angles', args.get('face_detector_angles'))
-	apply_state_item('face_detector_score', args.get('face_detector_score'))
-	# face landmarker
-	apply_state_item('face_landmarker_model', args.get('face_landmarker_model'))
-	apply_state_item('face_landmarker_score', args.get('face_landmarker_score'))
-	# face selector
-	apply_state_item('face_selector_mode', args.get('face_selector_mode'))
-	apply_state_item('face_selector_order', args.get('face_selector_order'))
-	apply_state_item('face_selector_age_start', args.get('face_selector_age_start'))
-	apply_state_item('face_selector_age_end', args.get('face_selector_age_end'))
-	apply_state_item('face_selector_gender', args.get('face_selector_gender'))
-	apply_state_item('face_selector_race', args.get('face_selector_race'))
-	apply_state_item('reference_face_position', args.get('reference_face_position'))
-	apply_state_item('reference_face_distance', args.get('reference_face_distance'))
-	apply_state_item('reference_frame_number', args.get('reference_frame_number'))
-	# face masker
-	apply_state_item('face_occluder_model', args.get('face_occluder_model'))
-	apply_state_item('face_parser_model', args.get('face_parser_model'))
-	apply_state_item('face_mask_types', args.get('face_mask_types'))
-	apply_state_item('face_mask_blur', args.get('face_mask_blur'))
-	apply_state_item('face_mask_padding', normalize_padding(args.get('face_mask_padding')))
-	apply_state_item('face_mask_regions', args.get('face_mask_regions'))
-	# frame extraction
-	apply_state_item('trim_frame_start', args.get('trim_frame_start'))
-	apply_state_item('trim_frame_end', args.get('trim_frame_end'))
-	apply_state_item('temp_frame_format', args.get('temp_frame_format'))
-	apply_state_item('keep_temp', args.get('keep_temp'))
-	# output creation
-	apply_state_item('output_image_quality', args.get('output_image_quality'))
-	if is_image(args.get('target_path')):
-		output_image_resolution = detect_image_resolution(args.get('target_path'))
-		output_image_resolutions = create_image_resolutions(output_image_resolution)
-		if args.get('output_image_resolution') in output_image_resolutions:
-			apply_state_item('output_image_resolution', args.get('output_image_resolution'))
-		else:
-			apply_state_item('output_image_resolution', pack_resolution(output_image_resolution))
-	apply_state_item('output_audio_encoder', args.get('output_audio_encoder'))
-	apply_state_item('output_audio_quality', args.get('output_audio_quality'))
-	apply_state_item('output_audio_volume', args.get('output_audio_volume'))
-	apply_state_item('output_video_encoder', args.get('output_video_encoder'))
-	apply_state_item('output_video_preset', args.get('output_video_preset'))
-	apply_state_item('output_video_quality', args.get('output_video_quality'))
-	if is_video(args.get('target_path')):
-		output_video_resolution = detect_video_resolution(args.get('target_path'))
-		output_video_resolutions = create_video_resolutions(output_video_resolution)
-		if args.get('output_video_resolution') in output_video_resolutions:
-			apply_state_item('output_video_resolution', args.get('output_video_resolution'))
-		else:
-			apply_state_item('output_video_resolution', pack_resolution(output_video_resolution))
-	if args.get('output_video_fps') or is_video(args.get('target_path')):
-		output_video_fps = normalize_fps(args.get('output_video_fps')) or detect_video_fps(args.get('target_path'))
-		apply_state_item('output_video_fps', output_video_fps)
-	# processors
-	available_processors = [ get_file_name(file_path) for file_path in resolve_file_paths('facefusion/processors/modules') ]
-	apply_state_item('processors', args.get('processors'))
-	for processor_module in get_processors_modules(available_processors):
-		processor_module.apply_args(args, apply_state_item)
-	# uis
-	apply_state_item('open_browser', args.get('open_browser'))
-	apply_state_item('ui_layouts', args.get('ui_layouts'))
-	apply_state_item('ui_workflow', args.get('ui_workflow'))
-	# execution
-	apply_state_item('execution_device_id', args.get('execution_device_id'))
-	apply_state_item('execution_providers', args.get('execution_providers'))
-	apply_state_item('execution_thread_count', args.get('execution_thread_count'))
-	apply_state_item('execution_queue_count', args.get('execution_queue_count'))
-	# download
-	apply_state_item('download_providers', args.get('download_providers'))
-	apply_state_item('download_scope', args.get('download_scope'))
-	# memory
-	apply_state_item('video_memory_strategy', args.get('video_memory_strategy'))
-	apply_state_item('system_memory_limit', args.get('system_memory_limit'))
-	# misc
-	apply_state_item('log_level', args.get('log_level'))
-	apply_state_item('halt_on_error', args.get('halt_on_error'))
-	# jobs
-	apply_state_item('job_id', args.get('job_id'))
-	apply_state_item('job_status', args.get('job_status'))
-	apply_state_item('step_index', args.get('step_index'))

facefusion/audio.py

@@ -5,13 +5,13 @@ import numpy
 import scipy
 from numpy.typing import NDArray
 
-from facefusion.ffmpeg import read_audio_buffer
+from facefusion import ffmpeg
 from facefusion.filesystem import is_audio
 from facefusion.types import Audio, AudioFrame, Fps, Mel, MelFilterBank, Spectrogram
 from facefusion.voice_extractor import batch_extract_voice
 
 
-@lru_cache()
+@lru_cache(maxsize = 64)
 def read_static_audio(audio_path : str, fps : Fps) -> Optional[List[AudioFrame]]:
 	return read_audio(audio_path, fps)
 
@@ -22,7 +22,7 @@ def read_audio(audio_path : str, fps : Fps) -> Optional[List[AudioFrame]]:
 	audio_channel_total = 2
 
 	if is_audio(audio_path):
-		audio_buffer = read_audio_buffer(audio_path, audio_sample_rate, audio_sample_size, audio_channel_total)
+		audio_buffer = ffmpeg.read_audio_buffer(audio_path, audio_sample_rate, audio_sample_size, audio_channel_total)
 		audio = numpy.frombuffer(audio_buffer, dtype = numpy.int16).reshape(-1, 2)
 		audio = prepare_audio(audio)
 		spectrogram = create_spectrogram(audio)
@@ -31,7 +31,7 @@ def read_audio(audio_path : str, fps : Fps) -> Optional[List[AudioFrame]]:
 	return None
 
 
-@lru_cache()
+@lru_cache(maxsize = 64)
 def read_static_voice(audio_path : str, fps : Fps) -> Optional[List[AudioFrame]]:
 	return read_voice(audio_path, fps)
 
@@ -44,7 +44,7 @@ def read_voice(audio_path : str, fps : Fps) -> Optional[List[AudioFrame]]:
 	voice_step_size = 180 * 1024
 
 	if is_audio(audio_path):
-		audio_buffer = read_audio_buffer(audio_path, voice_sample_rate, voice_sample_size, voice_channel_total)
+		audio_buffer = ffmpeg.read_audio_buffer(audio_path, voice_sample_rate, voice_sample_size, voice_channel_total)
 		audio = numpy.frombuffer(audio_buffer, dtype = numpy.int16).reshape(-1, 2)
 		audio = batch_extract_voice(audio, voice_chunk_size, voice_step_size)
 		audio = prepare_voice(audio)
@@ -62,7 +62,7 @@ def get_audio_frame(audio_path : str, fps : Fps, frame_number : int = 0) -> Opti
 	return None
 
 
-def extract_audio_frames(spectrogram: Spectrogram, fps: Fps) -> List[AudioFrame]:
+def extract_audio_frames(spectrogram : Spectrogram, fps : Fps) -> List[AudioFrame]:
 	audio_frames = []
 	mel_filter_total = 80
 	audio_step_size = 16
@@ -118,12 +118,12 @@ def convert_mel_to_hertz(mel : Mel) -> NDArray[Any]:
 
 def create_mel_filter_bank() -> MelFilterBank:
 	audio_sample_rate = 16000
-	audio_min_frequency = 55.0
-	audio_max_frequency = 7600.0
+	audio_frequency_min = 55.0
+	audio_frequency_max = 7600.0
 	mel_filter_total = 80
 	mel_bin_total = 800
 	mel_filter_bank = numpy.zeros((mel_filter_total, mel_bin_total // 2 + 1))
-	mel_frequency_range = numpy.linspace(convert_hertz_to_mel(audio_min_frequency), convert_hertz_to_mel(audio_max_frequency), mel_filter_total + 2)
+	mel_frequency_range = numpy.linspace(convert_hertz_to_mel(audio_frequency_min), convert_hertz_to_mel(audio_frequency_max), mel_filter_total + 2)
 	indices = numpy.floor((mel_bin_total + 1) * convert_mel_to_hertz(mel_frequency_range) / audio_sample_rate).astype(numpy.int16)
 
 	for index in range(mel_filter_total):

facefusion/benchmarker.py

@@ -0,0 +1,111 @@
+import hashlib
+import os
+import statistics
+import tempfile
+from time import perf_counter
+from typing import Iterator, List
+
+import facefusion.choices
+from facefusion import content_analyser, core, state_manager
+from facefusion.cli_helper import render_table
+from facefusion.download import conditional_download, resolve_download_url
+from facefusion.face_store import clear_faces
+from facefusion.filesystem import get_file_extension
+from facefusion.types import BenchmarkCycleSet
+from facefusion.vision import count_video_frame_total, detect_video_fps
+
+
+def pre_check() -> bool:
+	conditional_download('.assets/examples',
+	[
+		resolve_download_url('examples-3.0.0', 'source.jpg'),
+		resolve_download_url('examples-3.0.0', 'source.mp3'),
+		resolve_download_url('examples-3.0.0', 'target-240p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-360p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-540p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-720p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-1080p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-1440p.mp4'),
+		resolve_download_url('examples-3.0.0', 'target-2160p.mp4')
+	])
+	return True
+
+
+def run() -> Iterator[List[BenchmarkCycleSet]]:
+	benchmark_resolutions = state_manager.get_item('benchmark_resolutions')
+	benchmark_cycle_count = state_manager.get_item('benchmark_cycle_count')
+
+	state_manager.init_item('source_paths', [ '.assets/examples/source.jpg', '.assets/examples/source.mp3' ])
+	state_manager.init_item('face_landmarker_score', 0)
+	state_manager.init_item('temp_frame_format', 'bmp')
+	state_manager.init_item('output_audio_volume', 0)
+	state_manager.init_item('output_video_preset', 'ultrafast')
+	state_manager.init_item('video_memory_strategy', 'tolerant')
+
+	benchmarks = []
+	target_paths = [ facefusion.choices.benchmark_set.get(benchmark_resolution) for benchmark_resolution in benchmark_resolutions if benchmark_resolution in facefusion.choices.benchmark_set ]
+
+	for target_path in target_paths:
+		state_manager.init_item('target_path', target_path)
+		state_manager.init_item('output_path', suggest_output_path(state_manager.get_item('target_path')))
+		benchmarks.append(cycle(benchmark_cycle_count))
+		yield benchmarks
+
+
+def cycle(cycle_count : int) -> BenchmarkCycleSet:
+	process_times = []
+	video_frame_total = count_video_frame_total(state_manager.get_item('target_path'))
+	state_manager.init_item('output_video_fps', detect_video_fps(state_manager.get_item('target_path')))
+
+	if state_manager.get_item('benchmark_mode') == 'warm':
+		core.conditional_process()
+
+	for index in range(cycle_count):
+		if state_manager.get_item('benchmark_mode') == 'cold':
+			content_analyser.analyse_image.cache_clear()
+			content_analyser.analyse_video.cache_clear()
+			clear_faces()
+
+		start_time = perf_counter()
+		core.conditional_process()
+		end_time = perf_counter()
+		process_times.append(end_time - start_time)
+
+	average_run = round(statistics.mean(process_times), 2)
+	fastest_run = round(min(process_times), 2)
+	slowest_run = round(max(process_times), 2)
+	relative_fps = round(video_frame_total * cycle_count / sum(process_times), 2)
+
+	return\
+	{
+		'target_path': state_manager.get_item('target_path'),
+		'cycle_count': cycle_count,
+		'average_run': average_run,
+		'fastest_run': fastest_run,
+		'slowest_run': slowest_run,
+		'relative_fps': relative_fps
+	}
+
+
+def suggest_output_path(target_path : str) -> str:
+	target_file_extension = get_file_extension(target_path)
+	return os.path.join(tempfile.gettempdir(), hashlib.sha1(target_path.encode()).hexdigest() + target_file_extension)
+
+
+def render() -> None:
+	benchmarks = []
+	headers =\
+	[
+		'target_path',
+		'cycle_count',
+		'average_run',
+		'fastest_run',
+		'slowest_run',
+		'relative_fps'
+	]
+
+	for benchmark in run():
+		benchmarks = benchmark
+
+	contents = [ list(benchmark_set.values()) for benchmark_set in benchmarks ]
+	render_table(headers, contents)

facefusion/camera_manager.py

@@ -0,0 +1,53 @@
+from typing import List
+
+import cv2
+
+from facefusion.types import CameraPoolSet
+
+CAMERA_POOL_SET : CameraPoolSet =\
+{
+	'capture': {}
+}
+
+
+def get_local_camera_capture(camera_id : int) -> cv2.VideoCapture:
+	camera_key = str(camera_id)
+
+	if camera_key not in CAMERA_POOL_SET.get('capture'):
+		camera_capture = cv2.VideoCapture(camera_id)
+
+		if camera_capture.isOpened():
+			CAMERA_POOL_SET['capture'][camera_key] = camera_capture
+
+	return CAMERA_POOL_SET.get('capture').get(camera_key)
+
+
+def get_remote_camera_capture(camera_url : str) -> cv2.VideoCapture:
+	if camera_url not in CAMERA_POOL_SET.get('capture'):
+		camera_capture = cv2.VideoCapture(camera_url)
+
+		if camera_capture.isOpened():
+			CAMERA_POOL_SET['capture'][camera_url] = camera_capture
+
+	return CAMERA_POOL_SET.get('capture').get(camera_url)
+
+
+def clear_camera_pool() -> None:
+	for camera_capture in CAMERA_POOL_SET.get('capture').values():
+		camera_capture.release()
+
+	CAMERA_POOL_SET['capture'].clear()
+
+
+def detect_local_camera_ids(id_start : int, id_end : int) -> List[int]:
+	local_camera_ids = []
+
+	for camera_id in range(id_start, id_end):
+		cv2.utils.logging.setLogLevel(0)
+		camera_capture = get_local_camera_capture(camera_id)
+		cv2.utils.logging.setLogLevel(3)
+
+		if camera_capture and camera_capture.isOpened():
+			local_camera_ids.append(camera_id)
+
+	return local_camera_ids

facefusion/choices.py

@@ -1,25 +1,34 @@
 import logging
-from typing import List, Sequence
+from typing import List, Sequence, get_args
 
 from facefusion.common_helper import create_float_range, create_int_range
-from facefusion.types import Angle, AudioEncoder, AudioFormat, AudioTypeSet, DownloadProvider, DownloadProviderSet, DownloadScope, EncoderSet, ExecutionProvider, ExecutionProviderSet, FaceDetectorModel, FaceDetectorSet, FaceLandmarkerModel, FaceMaskRegion, FaceMaskRegionSet, FaceMaskType, FaceOccluderModel, FaceParserModel, FaceSelectorMode, FaceSelectorOrder, Gender, ImageFormat, ImageTypeSet, JobStatus, LogLevel, LogLevelSet, Race, Score, TempFrameFormat, UiWorkflow, VideoEncoder, VideoFormat, VideoMemoryStrategy, VideoPreset, VideoTypeSet, WebcamMode
+from facefusion.types import Angle, AudioEncoder, AudioFormat, AudioTypeSet, BenchmarkMode, BenchmarkResolution, BenchmarkSet, DownloadProvider, DownloadProviderSet, DownloadScope, EncoderSet, ExecutionProvider, ExecutionProviderSet, FaceDetectorModel, FaceDetectorSet, FaceLandmarkerModel, FaceMaskArea, FaceMaskAreaSet, FaceMaskRegion, FaceMaskRegionSet, FaceMaskType, FaceOccluderModel, FaceParserModel, FaceSelectorGender, FaceSelectorMode, FaceSelectorOrder, FaceSelectorRace, Gender, ImageFormat, ImageTypeSet, JobStatus, LogLevel, LogLevelSet, Race, Score, TempFrameFormat, TempPixelFormat, UiWorkflow, VideoEncoder, VideoFormat, VideoMemoryStrategy, VideoPreset, VideoTypeSet, VoiceExtractorModel, WorkflowMode, WorkflowStrategy
 
 face_detector_set : FaceDetectorSet =\
 {
 	'many': [ '640x640' ],
 	'retinaface': [ '160x160', '320x320', '480x480', '512x512', '640x640' ],
 	'scrfd': [ '160x160', '320x320', '480x480', '512x512', '640x640' ],
-	'yolo_face': [ '640x640' ]
+	'yolo_face': [ '640x640' ],
+	'yunet': [ '640x640' ]
+}
+face_detector_models : List[FaceDetectorModel] = list(get_args(FaceDetectorModel))
+face_landmarker_models : List[FaceLandmarkerModel] = list(get_args(FaceLandmarkerModel))
+face_selector_modes : List[FaceSelectorMode] = list(get_args(FaceSelectorMode))
+face_selector_orders : List[FaceSelectorOrder] = list(get_args(FaceSelectorOrder))
+genders : List[Gender] = list(get_args(Gender))
+races : List[Race] = list(get_args(Race))
+face_selector_genders : List[FaceSelectorGender] = list(get_args(FaceSelectorGender))
+face_selector_races : List[FaceSelectorRace] = list(get_args(FaceSelectorRace))
+face_occluder_models : List[FaceOccluderModel] = list(get_args(FaceOccluderModel))
+face_parser_models : List[FaceParserModel] = list(get_args(FaceParserModel))
+face_mask_types : List[FaceMaskType] = list(get_args(FaceMaskType))
+face_mask_area_set : FaceMaskAreaSet =\
+{
+	'upper-face': [ 0, 1, 2, 31, 32, 33, 34, 35, 14, 15, 16, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17 ],
+	'lower-face': [ 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 35, 34, 33, 32, 31 ],
+	'mouth': [ 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67 ]
 }
-face_detector_models : List[FaceDetectorModel] = list(face_detector_set.keys())
-face_landmarker_models : List[FaceLandmarkerModel] = [ 'many', '2dfan4', 'peppa_wutz' ]
-face_selector_modes : List[FaceSelectorMode] = [ 'many', 'one', 'reference' ]
-face_selector_orders : List[FaceSelectorOrder] = [ 'left-right', 'right-left', 'top-bottom', 'bottom-top', 'small-large', 'large-small', 'best-worst', 'worst-best' ]
-face_selector_genders : List[Gender] = [ 'female', 'male' ]
-face_selector_races : List[Race] = [ 'white', 'black', 'latino', 'asian', 'indian', 'arabic' ]
-face_occluder_models : List[FaceOccluderModel] = [ 'xseg_1', 'xseg_2', 'xseg_3' ]
-face_parser_models : List[FaceParserModel] = [ 'bisenet_resnet_18', 'bisenet_resnet_34' ]
-face_mask_types : List[FaceMaskType] = [ 'box', 'occlusion', 'region' ]
 face_mask_region_set : FaceMaskRegionSet =\
 {
 	'skin': 1,
@@ -33,7 +42,10 @@ face_mask_region_set : FaceMaskRegionSet =\
 	'upper-lip': 12,
 	'lower-lip': 13
 }
-face_mask_regions : List[FaceMaskRegion] = list(face_mask_region_set.keys())
+face_mask_areas : List[FaceMaskArea] = list(get_args(FaceMaskArea))
+face_mask_regions : List[FaceMaskRegion] = list(get_args(FaceMaskRegion))
+
+voice_extractor_models : List[VoiceExtractorModel] = list(get_args(VoiceExtractorModel))
 
 audio_type_set : AudioTypeSet =\
 {
@@ -58,40 +70,56 @@ video_type_set : VideoTypeSet =\
 	'm4v': 'video/mp4',
 	'mkv': 'video/x-matroska',
 	'mp4': 'video/mp4',
+	'mpeg': 'video/mpeg',
 	'mov': 'video/quicktime',
-	'webm': 'video/webm'
+	'mxf': 'application/mxf',
+	'webm': 'video/webm',
+	'wmv': 'video/x-ms-wmv'
 }
-audio_formats : List[AudioFormat] = list(audio_type_set.keys())
-image_formats : List[ImageFormat] = list(image_type_set.keys())
-video_formats : List[VideoFormat] = list(video_type_set.keys())
-temp_frame_formats : List[TempFrameFormat] = [ 'bmp', 'jpeg', 'png', 'tiff' ]
+workflow_modes : List[WorkflowMode] = list(get_args(WorkflowMode))
+workflow_strategies : List[WorkflowStrategy] = list(get_args(WorkflowStrategy))
 
+audio_formats : List[AudioFormat] = list(get_args(AudioFormat))
+image_formats : List[ImageFormat] = list(get_args(ImageFormat))
+video_formats : List[VideoFormat] = list(get_args(VideoFormat))
+temp_frame_formats : List[TempFrameFormat] = list(get_args(TempFrameFormat))
+temp_pixel_formats : List[TempPixelFormat] = list(get_args(TempPixelFormat))
+
+output_audio_encoders : List[AudioEncoder] = list(get_args(AudioEncoder))
+output_video_encoders : List[VideoEncoder] = list(get_args(VideoEncoder))
 output_encoder_set : EncoderSet =\
 {
-	'audio': [ 'flac', 'aac', 'libmp3lame', 'libopus', 'libvorbis', 'pcm_s16le', 'pcm_s32le' ],
-	'video': [ 'libx264', 'libx265', 'libvpx-vp9', 'h264_nvenc', 'hevc_nvenc', 'h264_amf', 'hevc_amf', 'h264_qsv', 'hevc_qsv', 'h264_videotoolbox', 'hevc_videotoolbox', 'rawvideo' ]
+	'audio': output_audio_encoders,
+	'video': output_video_encoders
 }
-output_audio_encoders : List[AudioEncoder] = output_encoder_set.get('audio')
-output_video_encoders : List[VideoEncoder] = output_encoder_set.get('video')
-output_video_presets : List[VideoPreset] = [ 'ultrafast', 'superfast', 'veryfast', 'faster', 'fast', 'medium', 'slow', 'slower', 'veryslow' ]
+output_video_presets : List[VideoPreset] = list(get_args(VideoPreset))
 
-image_template_sizes : List[float] = [ 0.25, 0.5, 0.75, 1, 1.5, 2, 2.5, 3, 3.5, 4 ]
-video_template_sizes : List[int] = [ 240, 360, 480, 540, 720, 1080, 1440, 2160, 4320 ]
-
-webcam_modes : List[WebcamMode] = [ 'inline', 'udp', 'v4l2' ]
-webcam_resolutions : List[str] = [ '320x240', '640x480', '800x600', '1024x768', '1280x720', '1280x960', '1920x1080', '2560x1440', '3840x2160' ]
+benchmark_modes : List[BenchmarkMode] = list(get_args(BenchmarkMode))
+benchmark_set : BenchmarkSet =\
+{
+	'240p': '.assets/examples/target-240p.mp4',
+	'360p': '.assets/examples/target-360p.mp4',
+	'540p': '.assets/examples/target-540p.mp4',
+	'720p': '.assets/examples/target-720p.mp4',
+	'1080p': '.assets/examples/target-1080p.mp4',
+	'1440p': '.assets/examples/target-1440p.mp4',
+	'2160p': '.assets/examples/target-2160p.mp4'
+}
+benchmark_resolutions : List[BenchmarkResolution] = list(get_args(BenchmarkResolution))
 
 execution_provider_set : ExecutionProviderSet =\
 {
 	'cuda': 'CUDAExecutionProvider',
 	'tensorrt': 'TensorrtExecutionProvider',
-	'directml': 'DmlExecutionProvider',
 	'rocm': 'ROCMExecutionProvider',
-	'openvino': 'OpenVINOExecutionProvider',
+	'migraphx': 'MIGraphXExecutionProvider',
 	'coreml': 'CoreMLExecutionProvider',
+	'openvino': 'OpenVINOExecutionProvider',
+	'qnn': 'QNNExecutionProvider',
+	'directml': 'DmlExecutionProvider',
 	'cpu': 'CPUExecutionProvider'
 }
-execution_providers : List[ExecutionProvider] = list(execution_provider_set.keys())
+execution_providers : List[ExecutionProvider] = list(get_args(ExecutionProvider))
 download_provider_set : DownloadProviderSet =\
 {
 	'github':
@@ -112,10 +140,10 @@ download_provider_set : DownloadProviderSet =\
 		'path': '/facefusion/{base_name}/resolve/main/{file_name}'
 	}
 }
-download_providers : List[DownloadProvider] = list(download_provider_set.keys())
-download_scopes : List[DownloadScope] = [ 'lite', 'full' ]
+download_providers : List[DownloadProvider] = list(get_args(DownloadProvider))
+download_scopes : List[DownloadScope] = list(get_args(DownloadScope))
 
-video_memory_strategies : List[VideoMemoryStrategy] = [ 'strict', 'moderate', 'tolerant' ]
+video_memory_strategies : List[VideoMemoryStrategy] = list(get_args(VideoMemoryStrategy))
 
 log_level_set : LogLevelSet =\
 {
@@ -124,14 +152,14 @@ log_level_set : LogLevelSet =\
 	'info': logging.INFO,
 	'debug': logging.DEBUG
 }
-log_levels : List[LogLevel] = list(log_level_set.keys())
+log_levels : List[LogLevel] = list(get_args(LogLevel))
 
-ui_workflows : List[UiWorkflow] = [ 'instant_runner', 'job_runner', 'job_manager' ]
-job_statuses : List[JobStatus] = [ 'drafted', 'queued', 'completed', 'failed' ]
+ui_workflows : List[UiWorkflow] = list(get_args(UiWorkflow))
+job_statuses : List[JobStatus] = list(get_args(JobStatus))
 
+benchmark_cycle_count_range : Sequence[int] = create_int_range(1, 10, 1)
 execution_thread_count_range : Sequence[int] = create_int_range(1, 32, 1)
-execution_queue_count_range : Sequence[int] = create_int_range(1, 4, 1)
-system_memory_limit_range : Sequence[int] = create_int_range(0, 128, 4)
+face_detector_margin_range : Sequence[int] = create_int_range(0, 100, 1)
 face_detector_angles : Sequence[Angle] = create_int_range(0, 270, 90)
 face_detector_score_range : Sequence[Score] = create_float_range(0.0, 1.0, 0.05)
 face_landmarker_score_range : Sequence[Score] = create_float_range(0.0, 1.0, 0.05)
@@ -139,7 +167,11 @@ face_mask_blur_range : Sequence[float] = create_float_range(0.0, 1.0, 0.05)
 face_mask_padding_range : Sequence[int] = create_int_range(0, 100, 1)
 face_selector_age_range : Sequence[int] = create_int_range(0, 100, 1)
 reference_face_distance_range : Sequence[float] = create_float_range(0.0, 1.0, 0.05)
+face_tracker_score_range : Sequence[Score] = create_float_range(0.0, 0.5, 0.05)
+target_frame_amount_range : Sequence[int] = create_int_range(0, 10, 1)
 output_image_quality_range : Sequence[int] = create_int_range(0, 100, 1)
+output_image_scale_range : Sequence[float] = create_float_range(0.25, 8.0, 0.25)
 output_audio_quality_range : Sequence[int] = create_int_range(0, 100, 1)
 output_audio_volume_range : Sequence[int] = create_int_range(0, 100, 1)
 output_video_quality_range : Sequence[int] = create_int_range(0, 100, 1)
+output_video_scale_range : Sequence[float] = create_float_range(0.25, 8.0, 0.25)

facefusion/cli_helper.py

@@ -1,25 +1,25 @@
-from typing import Tuple
+from typing import List, Tuple
 
 from facefusion.logger import get_package_logger
-from facefusion.types import TableContents, TableHeaders
+from facefusion.types import TableContent, TableHeader
 
 
-def render_table(headers : TableHeaders, contents : TableContents) -> None:
+def render_table(headers : List[TableHeader], contents : List[List[TableContent]]) -> None:
 	package_logger = get_package_logger()
 	table_column, table_separator = create_table_parts(headers, contents)
 
-	package_logger.info(table_separator)
-	package_logger.info(table_column.format(*headers))
-	package_logger.info(table_separator)
+	package_logger.critical(table_separator)
+	package_logger.critical(table_column.format(*headers))
+	package_logger.critical(table_separator)
 
 	for content in contents:
-		content = [ value if value else '' for value in content ]
-		package_logger.info(table_column.format(*content))
+		content = [ str(value) for value in content ]
+		package_logger.critical(table_column.format(*content))
 
-	package_logger.info(table_separator)
+	package_logger.critical(table_separator)
 
 
-def create_table_parts(headers : TableHeaders, contents : TableContents) -> Tuple[str, str]:
+def create_table_parts(headers : List[TableHeader], contents : List[List[TableContent]]) -> Tuple[str, str]:
 	column_parts = []
 	separator_parts = []
 	widths = [ len(header) for header in headers ]

facefusion/common_helper.py

@@ -15,11 +15,11 @@ def is_windows() -> bool:
 
 
 def create_int_metavar(int_range : Sequence[int]) -> str:
-	return '[' + str(int_range[0]) + '..' + str(int_range[-1]) + ':' + str(calc_int_step(int_range)) + ']'
+	return '[' + str(int_range[0]) + '..' + str(int_range[-1]) + ':' + str(calculate_int_step(int_range)) + ']'
 
 
 def create_float_metavar(float_range : Sequence[float]) -> str:
-	return '[' + str(float_range[0]) + '..' + str(float_range[-1]) + ':' + str(calc_float_step(float_range)) + ']'
+	return '[' + str(float_range[0]) + '..' + str(float_range[-1]) + ':' + str(calculate_float_step(float_range)) + ']'
 
 
 def create_int_range(start : int, end : int, step : int) -> Sequence[int]:
@@ -42,11 +42,11 @@ def create_float_range(start : float, end : float, step : float) -> Sequence[flo
 	return float_range
 
 
-def calc_int_step(int_range : Sequence[int]) -> int:
+def calculate_int_step(int_range : Sequence[int]) -> int:
 	return int_range[1] - int_range[0]
 
 
-def calc_float_step(float_range : Sequence[float]) -> float:
+def calculate_float_step(float_range : Sequence[float]) -> float:
 	return round(float_range[1] - float_range[0], 2)
 
 
@@ -78,6 +78,12 @@ def get_first(__list__ : Any) -> Any:
 	return None
 
 
+def get_middle(__list__ : Any) -> Any:
+	if isinstance(__list__, Sequence) and __list__:
+		return __list__[len(__list__) // 2]
+	return None
+
+
 def get_last(__list__ : Any) -> Any:
 	if isinstance(__list__, Reversible):
 		return next(reversed(__list__), None)

facefusion/conda.py

@@ -0,0 +1,41 @@
+import os
+import sys
+from typing import List
+
+from facefusion.common_helper import is_linux, is_windows
+
+
+def setup() -> None:
+	conda_prefix = os.getenv('CONDA_PREFIX')
+	conda_ready = os.getenv('CONDA_READY')
+
+	if conda_prefix and not conda_ready:
+		if is_linux():
+			python_id = 'python' + str(sys.version_info.major) + '.' + str(sys.version_info.minor)
+			library_paths : List[str] =\
+			[
+				os.path.join(conda_prefix, 'lib'),
+				os.path.join(conda_prefix, 'lib', python_id, 'site-packages', 'tensorrt_libs')
+			]
+			library_paths = list(filter(os.path.exists, library_paths))
+
+			if library_paths:
+				if os.getenv('LD_LIBRARY_PATH'):
+					library_paths.append(os.getenv('LD_LIBRARY_PATH'))
+				os.environ['LD_LIBRARY_PATH'] = os.pathsep.join(library_paths)
+				os.environ['CONDA_READY'] = '1'
+				os.execv(sys.executable, [ sys.executable ] + sys.argv)
+
+		if is_windows():
+			library_paths =\
+			[
+				os.path.join(conda_prefix, 'Lib'),
+				os.path.join(conda_prefix, 'Lib', 'site-packages', 'tensorrt_libs')
+			]
+			library_paths = list(filter(os.path.exists, library_paths))
+
+			if library_paths:
+				if os.getenv('PATH'):
+					library_paths.append(os.getenv('PATH'))
+				os.environ['PATH'] = os.pathsep.join(library_paths)
+				os.environ['CONDA_READY'] = '1'

facefusion/config.py

@@ -1,29 +1,20 @@
 from configparser import ConfigParser
+from functools import lru_cache
 from typing import List, Optional
 
 from facefusion import state_manager
 from facefusion.common_helper import cast_bool, cast_float, cast_int
 
-CONFIG_PARSER = None
 
-
-def get_config_parser() -> ConfigParser:
-	global CONFIG_PARSER
-
-	if CONFIG_PARSER is None:
-		CONFIG_PARSER = ConfigParser()
-		CONFIG_PARSER.read(state_manager.get_item('config_path'), encoding = 'utf-8')
-	return CONFIG_PARSER
-
-
-def clear_config_parser() -> None:
-	global CONFIG_PARSER
-
-	CONFIG_PARSER = None
+@lru_cache
+def get_static_config_parser() -> ConfigParser:
+	config_parser = ConfigParser()
+	config_parser.read(state_manager.get_item('config_path'), encoding = 'utf-8')
+	return config_parser
 
 
 def get_str_value(section : str, option : str, fallback : Optional[str] = None) -> Optional[str]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return config_parser.get(section, option)
@@ -31,7 +22,7 @@ def get_str_value(section : str, option : str, fallback : Optional[str] = None)
 
 
 def get_int_value(section : str, option : str, fallback : Optional[str] = None) -> Optional[int]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return config_parser.getint(section, option)
@@ -39,7 +30,7 @@ def get_int_value(section : str, option : str, fallback : Optional[str] = None)
 
 
 def get_float_value(section : str, option : str, fallback : Optional[str] = None) -> Optional[float]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return config_parser.getfloat(section, option)
@@ -47,7 +38,7 @@ def get_float_value(section : str, option : str, fallback : Optional[str] = None
 
 
 def get_bool_value(section : str, option : str, fallback : Optional[str] = None) -> Optional[bool]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return config_parser.getboolean(section, option)
@@ -55,7 +46,7 @@ def get_bool_value(section : str, option : str, fallback : Optional[str] = None)
 
 
 def get_str_list(section : str, option : str, fallback : Optional[str] = None) -> Optional[List[str]]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return config_parser.get(section, option).split()
@@ -65,7 +56,7 @@ def get_str_list(section : str, option : str, fallback : Optional[str] = None) -
 
 
 def get_int_list(section : str, option : str, fallback : Optional[str] = None) -> Optional[List[int]]:
-	config_parser = get_config_parser()
+	config_parser = get_static_config_parser()
 
 	if config_parser.has_option(section, option) and config_parser.get(section, option).strip():
 		return list(map(int, config_parser.get(section, option).split()))

facefusion/content_analyser.py

@@ -1,65 +1,136 @@
 from functools import lru_cache
-from typing import List
+from typing import Tuple
 
 import numpy
 from tqdm import tqdm
 
-from facefusion import inference_manager, state_manager, wording
+from facefusion import inference_manager, state_manager, translator, video_manager
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
 from facefusion.filesystem import resolve_relative_path
 from facefusion.thread_helper import conditional_thread_semaphore
-from facefusion.types import Detection, DownloadScope, Fps, InferencePool, ModelOptions, ModelSet, Score, VisionFrame
-from facefusion.vision import detect_video_fps, fit_frame, read_image, read_video_frame
+from facefusion.types import Detection, DownloadScope, DownloadSet, Fps, InferencePool, ModelSet, VisionFrame
+from facefusion.vision import detect_video_fps, fit_contain_frame, read_image
 
 STREAM_COUNTER = 0
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
-		'yolo_nsfw':
+		'nsfw_1':
 		{
+			'__metadata__':
+			{
+				'vendor': 'EraX',
+				'license': 'Apache-2.0',
+				'year': 2024
+			},
 			'hashes':
 			{
 				'content_analyser':
 				{
-					'url': resolve_download_url('models-3.2.0', 'yolo_11m_nsfw.hash'),
-					'path': resolve_relative_path('../.assets/models/yolo_11m_nsfw.hash')
+					'url': resolve_download_url('models-3.3.0', 'nsfw_1.hash'),
+					'path': resolve_relative_path('../.assets/models/nsfw_1.hash')
 				}
 			},
 			'sources':
 			{
 				'content_analyser':
 				{
-					'url': resolve_download_url('models-3.2.0', 'yolo_11m_nsfw.onnx'),
-					'path': resolve_relative_path('../.assets/models/yolo_11m_nsfw.onnx')
+					'url': resolve_download_url('models-3.3.0', 'nsfw_1.onnx'),
+					'path': resolve_relative_path('../.assets/models/nsfw_1.onnx')
 				}
 			},
-			'size': (640, 640)
+			'size': (640, 640),
+			'mean': (0.0, 0.0, 0.0),
+			'standard_deviation': (1.0, 1.0, 1.0)
+		},
+		'nsfw_2':
+		{
+			'__metadata__':
+			{
+				'vendor': 'Marqo',
+				'license': 'Apache-2.0',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'content_analyser':
+				{
+					'url': resolve_download_url('models-3.3.0', 'nsfw_2.hash'),
+					'path': resolve_relative_path('../.assets/models/nsfw_2.hash')
+				}
+			},
+			'sources':
+			{
+				'content_analyser':
+				{
+					'url': resolve_download_url('models-3.3.0', 'nsfw_2.onnx'),
+					'path': resolve_relative_path('../.assets/models/nsfw_2.onnx')
+				}
+			},
+			'size': (384, 384),
+			'mean': (0.5, 0.5, 0.5),
+			'standard_deviation': (0.5, 0.5, 0.5)
+		},
+		'nsfw_3':
+		{
+			'__metadata__':
+			{
+				'vendor': 'Freepik',
+				'license': 'MIT',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'content_analyser':
+				{
+					'url': resolve_download_url('models-3.3.0', 'nsfw_3.hash'),
+					'path': resolve_relative_path('../.assets/models/nsfw_3.hash')
+				}
+			},
+			'sources':
+			{
+				'content_analyser':
+				{
+					'url': resolve_download_url('models-3.3.0', 'nsfw_3.onnx'),
+					'path': resolve_relative_path('../.assets/models/nsfw_3.onnx')
+				}
+			},
+			'size': (448, 448),
+			'mean': (0.48145466, 0.4578275, 0.40821073),
+			'standard_deviation': (0.26862954, 0.26130258, 0.27577711)
 		}
 	}
 
 
 def get_inference_pool() -> InferencePool:
-	model_names = [ 'yolo_nsfw' ]
-	model_source_set = get_model_options().get('sources')
+	model_names = [ 'nsfw_1', 'nsfw_2', 'nsfw_3' ]
+	_, model_source_set = collect_model_downloads()
 
 	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
 
 
 def clear_inference_pool() -> None:
-	model_names = [ 'yolo_nsfw' ]
+	model_names = [ 'nsfw_1', 'nsfw_2', 'nsfw_3' ]
 	inference_manager.clear_inference_pool(__name__, model_names)
 
 
-def get_model_options() -> ModelOptions:
-	return create_static_model_set('full').get('yolo_nsfw')
+def collect_model_downloads() -> Tuple[DownloadSet, DownloadSet]:
+	model_set = create_static_model_set('full')
+	model_hash_set = {}
+	model_source_set = {}
+
+	for content_analyser_model in [ 'nsfw_1', 'nsfw_2', 'nsfw_3' ]:
+		model_hash_set[content_analyser_model] = model_set.get(content_analyser_model).get('hashes').get('content_analyser')
+		model_source_set[content_analyser_model] = model_set.get(content_analyser_model).get('sources').get('content_analyser')
+
+	return model_hash_set, model_source_set
 
 
 def pre_check() -> bool:
-	model_hash_set = get_model_options().get('hashes')
-	model_source_set = get_model_options().get('sources')
+	model_hash_set, model_source_set = collect_model_downloads()
 
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
@@ -74,71 +145,101 @@ def analyse_stream(vision_frame : VisionFrame, video_fps : Fps) -> bool:
 
 
 def analyse_frame(vision_frame : VisionFrame) -> bool:
-	nsfw_scores = detect_nsfw(vision_frame)
-
-	return len(nsfw_scores) > 0
+	return detect_nsfw(vision_frame)
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def analyse_image(image_path : str) -> bool:
 	vision_frame = read_image(image_path)
 	return analyse_frame(vision_frame)
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def analyse_video(video_path : str, trim_frame_start : int, trim_frame_end : int) -> bool:
 	video_fps = detect_video_fps(video_path)
 	frame_range = range(trim_frame_start, trim_frame_end)
+	video_reader = video_manager.get_reader(video_path, 'analyse_video')
 	rate = 0.0
 	total = 0
 	counter = 0
 
-	with tqdm(total = len(frame_range), desc = wording.get('analysing'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
+	if trim_frame_start > 0:
+		video_manager.seek_video_reader(video_reader, trim_frame_start)
+
+	with tqdm(total = len(frame_range), desc = translator.get('analysing'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
 
 		for frame_number in frame_range:
+			vision_frame = video_manager.read_video_frame(video_reader)
+
 			if frame_number % int(video_fps) == 0:
-				vision_frame = read_video_frame(video_path, frame_number)
-				total += 1
-				if analyse_frame(vision_frame):
-					counter += 1
+				if numpy.any(vision_frame):
+					total += 1
+
+					if analyse_frame(vision_frame):
+						counter += 1
+
 			if counter > 0 and total > 0:
 				rate = counter / total * 100
+
 			progress.set_postfix(rate = rate)
 			progress.update()
 
-	return rate > 10.0
+	return bool(rate > 10.0)
 
 
-def detect_nsfw(vision_frame : VisionFrame) -> List[Score]:
-	nsfw_scores = []
-	model_size = get_model_options().get('size')
-	temp_vision_frame = fit_frame(vision_frame, model_size)
-	detect_vision_frame = prepare_detect_frame(temp_vision_frame)
-	detection = forward(detect_vision_frame)
-	detection = numpy.squeeze(detection).T
-	nsfw_scores_raw = numpy.amax(detection[:, 4:], axis = 1)
-	keep_indices = numpy.where(nsfw_scores_raw > 0.2)[0]
+def detect_nsfw(vision_frame : VisionFrame) -> bool:
+	is_nsfw_1 = detect_with_nsfw_1(vision_frame)
+	is_nsfw_2 = detect_with_nsfw_2(vision_frame)
+	is_nsfw_3 = detect_with_nsfw_3(vision_frame)
 
-	if numpy.any(keep_indices):
-		nsfw_scores_raw = nsfw_scores_raw[keep_indices]
-		nsfw_scores = nsfw_scores_raw.ravel().tolist()
-
-	return nsfw_scores
+	return is_nsfw_1 and is_nsfw_2 or is_nsfw_1 and is_nsfw_3 or is_nsfw_2 and is_nsfw_3
 
 
-def forward(vision_frame : VisionFrame) -> Detection:
-	content_analyser = get_inference_pool().get('content_analyser')
+def detect_with_nsfw_1(vision_frame : VisionFrame) -> bool:
+	detect_vision_frame = prepare_detect_frame(vision_frame, 'nsfw_1')
+	detection = forward_nsfw(detect_vision_frame, 'nsfw_1')
+	detection_score = numpy.max(numpy.amax(detection[:, 4:], axis = 1))
+	return bool(detection_score > 0.2)
+
+
+def detect_with_nsfw_2(vision_frame : VisionFrame) -> bool:
+	detect_vision_frame = prepare_detect_frame(vision_frame, 'nsfw_2')
+	detection = forward_nsfw(detect_vision_frame, 'nsfw_2')
+	detection_score = detection[0] - detection[1]
+	return bool(detection_score > 0.25)
+
+
+def detect_with_nsfw_3(vision_frame : VisionFrame) -> bool:
+	detect_vision_frame = prepare_detect_frame(vision_frame, 'nsfw_3')
+	detection = forward_nsfw(detect_vision_frame, 'nsfw_3')
+	detection_score = (detection[2] + detection[3]) - (detection[0] + detection[1])
+	return bool(detection_score > 10.5)
+
+
+def forward_nsfw(vision_frame : VisionFrame, model_name : str) -> Detection:
+	content_analyser = get_inference_pool().get(model_name)
 
 	with conditional_thread_semaphore():
 		detection = content_analyser.run(None,
 		{
 			'input': vision_frame
-		})
+		})[0]
+
+	if model_name in [ 'nsfw_2', 'nsfw_3' ]:
+		return detection[0]
 
 	return detection
 
 
-def prepare_detect_frame(temp_vision_frame : VisionFrame) -> VisionFrame:
-	detect_vision_frame = temp_vision_frame / 255.0
+def prepare_detect_frame(temp_vision_frame : VisionFrame, model_name : str) -> VisionFrame:
+	model_set = create_static_model_set('full').get(model_name)
+	model_size = model_set.get('size')
+	model_mean = model_set.get('mean')
+	model_standard_deviation = model_set.get('standard_deviation')
+
+	detect_vision_frame = fit_contain_frame(temp_vision_frame, model_size)
+	detect_vision_frame = detect_vision_frame[:, :, ::-1] / 255.0
+	detect_vision_frame -= model_mean
+	detect_vision_frame /= model_standard_deviation
 	detect_vision_frame = numpy.expand_dims(detect_vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
 	return detect_vision_frame

facefusion/core.py

@@ -1,37 +1,27 @@
+import inspect
 import itertools
 import shutil
 import signal
 import sys
 from time import time
 
-import numpy
-
-from facefusion import cli_helper, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, process_manager, state_manager, voice_extractor, wording
+from facefusion import benchmarker, cli_helper, content_analyser, hash_helper, logger, state_manager, translator
 from facefusion.args import apply_args, collect_job_args, reduce_job_args, reduce_step_args
-from facefusion.common_helper import get_first
-from facefusion.content_analyser import analyse_image, analyse_video
 from facefusion.download import conditional_download_hashes, conditional_download_sources
-from facefusion.exit_helper import graceful_exit, hard_exit
-from facefusion.face_analyser import get_average_face, get_many_faces, get_one_face
-from facefusion.face_selector import sort_and_filter_faces
-from facefusion.face_store import append_reference_face, clear_reference_faces, get_reference_faces
-from facefusion.ffmpeg import copy_image, extract_frames, finalize_image, merge_video, replace_audio, restore_audio
-from facefusion.filesystem import filter_audio_paths, get_file_name, is_image, is_video, resolve_file_paths, resolve_file_pattern
+from facefusion.exit_helper import hard_exit, signal_exit
+from facefusion.filesystem import get_file_extension, get_file_name, is_video, resolve_file_paths, resolve_file_pattern
 from facefusion.jobs import job_helper, job_manager, job_runner
 from facefusion.jobs.job_list import compose_job_list
-from facefusion.memory import limit_system_memory
 from facefusion.processors.core import get_processors_modules
 from facefusion.program import create_program
 from facefusion.program_helper import validate_args
-from facefusion.statistics import conditional_log_statistics
-from facefusion.temp_helper import clear_temp_directory, create_temp_directory, get_temp_file_path, move_temp_file, resolve_temp_frame_paths
-from facefusion.types import Args, ErrorCode
-from facefusion.vision import pack_resolution, read_image, read_static_images, read_video_frame, restrict_image_resolution, restrict_trim_frame, restrict_video_fps, restrict_video_resolution, unpack_resolution
+from facefusion.types import Args, ErrorCode, WorkflowMode
+from facefusion.workflows import image_to_image, image_to_video
 
 
 def cli() -> None:
 	if pre_check():
-		signal.signal(signal.SIGINT, lambda signal_number, frame: graceful_exit(0))
+		signal.signal(signal.SIGINT, signal_exit)
 		program = create_program()
 
 		if validate_args(program):
@@ -50,14 +40,14 @@ def cli() -> None:
 
 
 def route(args : Args) -> None:
-	system_memory_limit = state_manager.get_item('system_memory_limit')
-
-	if system_memory_limit and system_memory_limit > 0:
-		limit_system_memory(system_memory_limit)
-
 	if state_manager.get_item('command') == 'force-download':
 		error_code = force_download()
-		return hard_exit(error_code)
+		hard_exit(error_code)
+
+	if state_manager.get_item('command') == 'benchmark':
+		if not common_pre_check() or not processors_pre_check() or not benchmarker.pre_check():
+			hard_exit(2)
+		benchmarker.render()
 
 	if state_manager.get_item('command') in [ 'job-list', 'job-create', 'job-submit', 'job-submit-all', 'job-delete', 'job-delete-all', 'job-add-step', 'job-remix-step', 'job-insert-step', 'job-remove-step' ]:
 		if not job_manager.init_jobs(state_manager.get_item('jobs_path')):
@@ -69,24 +59,24 @@ def route(args : Args) -> None:
 		import facefusion.uis.core as ui
 
 		if not common_pre_check() or not processors_pre_check():
-			return hard_exit(2)
+			hard_exit(2)
 		for ui_layout in ui.get_ui_layouts_modules(state_manager.get_item('ui_layouts')):
 			if not ui_layout.pre_check():
-				return hard_exit(2)
+				hard_exit(2)
 		ui.init()
 		ui.launch()
 
 	if state_manager.get_item('command') == 'headless-run':
 		if not job_manager.init_jobs(state_manager.get_item('jobs_path')):
 			hard_exit(1)
-		error_core = process_headless(args)
-		hard_exit(error_core)
+		error_code = process_headless(args)
+		hard_exit(error_code)
 
 	if state_manager.get_item('command') == 'batch-run':
 		if not job_manager.init_jobs(state_manager.get_item('jobs_path')):
 			hard_exit(1)
-		error_core = process_batch(args)
-		hard_exit(error_core)
+		error_code = process_batch(args)
+		hard_exit(error_code)
 
 	if state_manager.get_item('command') in [ 'job-run', 'job-run-all', 'job-retry', 'job-retry-all' ]:
 		if not job_manager.init_jobs(state_manager.get_item('jobs_path')):
@@ -97,32 +87,20 @@ def route(args : Args) -> None:
 
 def pre_check() -> bool:
 	if sys.version_info < (3, 10):
-		logger.error(wording.get('python_not_supported').format(version = '3.10'), __name__)
+		logger.error(translator.get('python_not_supported').format(version = '3.10'), __name__)
 		return False
 
-	if not shutil.which('curl'):
-		logger.error(wording.get('curl_not_installed'), __name__)
-		return False
-
-	if not shutil.which('ffmpeg'):
-		logger.error(wording.get('ffmpeg_not_installed'), __name__)
-		return False
+	for dependency in [ 'curl', 'ffmpeg', 'ffprobe' ]:
+		if not shutil.which(dependency):
+			logger.error(translator.get('dependency_not_installed').format(dependency = dependency), __name__)
+			return False
 	return True
 
 
 def common_pre_check() -> bool:
-	common_modules =\
-	[
-		content_analyser,
-		face_classifier,
-		face_detector,
-		face_landmarker,
-		face_masker,
-		face_recognizer,
-		voice_extractor
-	]
+	content_analyser_content = inspect.getsource(content_analyser).encode()
 
-	return all(module.pre_check() for module in common_modules)
+	return hash_helper.create_hash(content_analyser_content) == '0922c180'
 
 
 def processors_pre_check() -> bool:
@@ -133,22 +111,19 @@ def processors_pre_check() -> bool:
 
 
 def force_download() -> ErrorCode:
-	common_modules =\
-	[
-		content_analyser,
-		face_classifier,
-		face_detector,
-		face_landmarker,
-		face_masker,
-		face_recognizer,
-		voice_extractor
-	]
+	download_scope = state_manager.get_item('download_scope')
 	available_processors = [ get_file_name(file_path) for file_path in resolve_file_paths('facefusion/processors/modules') ]
 	processor_modules = get_processors_modules(available_processors)
+	common_modules = []
+
+	for processor_module in processor_modules:
+		for common_module in processor_module.get_common_modules():
+			if common_module not in common_modules:
+				common_modules.append(common_module)
 
 	for module in common_modules + processor_modules:
 		if hasattr(module, 'create_static_model_set'):
-			for model in module.create_static_model_set(state_manager.get_item('download_scope')).values():
+			for model in module.create_static_model_set(download_scope).values():
 				model_hash_set = model.get('hashes')
 				model_source_set = model.get('sources')
 
@@ -170,106 +145,106 @@ def route_job_manager(args : Args) -> ErrorCode:
 
 	if state_manager.get_item('command') == 'job-create':
 		if job_manager.create_job(state_manager.get_item('job_id')):
-			logger.info(wording.get('job_created').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('job_created').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.error(wording.get('job_not_created').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.error(translator.get('job_not_created').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-submit':
 		if job_manager.submit_job(state_manager.get_item('job_id')):
-			logger.info(wording.get('job_submitted').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('job_submitted').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.error(wording.get('job_not_submitted').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.error(translator.get('job_not_submitted').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-submit-all':
 		if job_manager.submit_jobs(state_manager.get_item('halt_on_error')):
-			logger.info(wording.get('job_all_submitted'), __name__)
+			logger.info(translator.get('job_all_submitted'), __name__)
 			return 0
-		logger.error(wording.get('job_all_not_submitted'), __name__)
+		logger.error(translator.get('job_all_not_submitted'), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-delete':
 		if job_manager.delete_job(state_manager.get_item('job_id')):
-			logger.info(wording.get('job_deleted').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('job_deleted').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.error(wording.get('job_not_deleted').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.error(translator.get('job_not_deleted').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-delete-all':
 		if job_manager.delete_jobs(state_manager.get_item('halt_on_error')):
-			logger.info(wording.get('job_all_deleted'), __name__)
+			logger.info(translator.get('job_all_deleted'), __name__)
 			return 0
-		logger.error(wording.get('job_all_not_deleted'), __name__)
+		logger.error(translator.get('job_all_not_deleted'), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-add-step':
 		step_args = reduce_step_args(args)
 
 		if job_manager.add_step(state_manager.get_item('job_id'), step_args):
-			logger.info(wording.get('job_step_added').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('job_step_added').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.error(wording.get('job_step_not_added').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.error(translator.get('job_step_not_added').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-remix-step':
 		step_args = reduce_step_args(args)
 
 		if job_manager.remix_step(state_manager.get_item('job_id'), state_manager.get_item('step_index'), step_args):
-			logger.info(wording.get('job_remix_step_added').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+			logger.info(translator.get('job_remix_step_added').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 			return 0
-		logger.error(wording.get('job_remix_step_not_added').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+		logger.error(translator.get('job_remix_step_not_added').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-insert-step':
 		step_args = reduce_step_args(args)
 
 		if job_manager.insert_step(state_manager.get_item('job_id'), state_manager.get_item('step_index'), step_args):
-			logger.info(wording.get('job_step_inserted').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+			logger.info(translator.get('job_step_inserted').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 			return 0
-		logger.error(wording.get('job_step_not_inserted').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+		logger.error(translator.get('job_step_not_inserted').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-remove-step':
 		if job_manager.remove_step(state_manager.get_item('job_id'), state_manager.get_item('step_index')):
-			logger.info(wording.get('job_step_removed').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+			logger.info(translator.get('job_step_removed').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 			return 0
-		logger.error(wording.get('job_step_not_removed').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
+		logger.error(translator.get('job_step_not_removed').format(job_id = state_manager.get_item('job_id'), step_index = state_manager.get_item('step_index')), __name__)
 		return 1
 	return 1
 
 
 def route_job_runner() -> ErrorCode:
 	if state_manager.get_item('command') == 'job-run':
-		logger.info(wording.get('running_job').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.info(translator.get('running_job').format(job_id = state_manager.get_item('job_id')), __name__)
 		if job_runner.run_job(state_manager.get_item('job_id'), process_step):
-			logger.info(wording.get('processing_job_succeed').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('processing_job_succeeded').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.info(wording.get('processing_job_failed').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.info(translator.get('processing_job_failed').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-run-all':
-		logger.info(wording.get('running_jobs'), __name__)
+		logger.info(translator.get('running_jobs'), __name__)
 		if job_runner.run_jobs(process_step, state_manager.get_item('halt_on_error')):
-			logger.info(wording.get('processing_jobs_succeed'), __name__)
+			logger.info(translator.get('processing_jobs_succeeded'), __name__)
 			return 0
-		logger.info(wording.get('processing_jobs_failed'), __name__)
+		logger.info(translator.get('processing_jobs_failed'), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-retry':
-		logger.info(wording.get('retrying_job').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.info(translator.get('retrying_job').format(job_id = state_manager.get_item('job_id')), __name__)
 		if job_runner.retry_job(state_manager.get_item('job_id'), process_step):
-			logger.info(wording.get('processing_job_succeed').format(job_id = state_manager.get_item('job_id')), __name__)
+			logger.info(translator.get('processing_job_succeeded').format(job_id = state_manager.get_item('job_id')), __name__)
 			return 0
-		logger.info(wording.get('processing_job_failed').format(job_id = state_manager.get_item('job_id')), __name__)
+		logger.info(translator.get('processing_job_failed').format(job_id = state_manager.get_item('job_id')), __name__)
 		return 1
 
 	if state_manager.get_item('command') == 'job-retry-all':
-		logger.info(wording.get('retrying_jobs'), __name__)
+		logger.info(translator.get('retrying_jobs'), __name__)
 		if job_runner.retry_jobs(process_step, state_manager.get_item('halt_on_error')):
-			logger.info(wording.get('processing_jobs_succeed'), __name__)
+			logger.info(translator.get('processing_jobs_succeeded'), __name__)
 			return 0
-		logger.info(wording.get('processing_jobs_failed'), __name__)
+		logger.info(translator.get('processing_jobs_failed'), __name__)
 		return 1
 	return 2
 
@@ -295,7 +270,12 @@ def process_batch(args : Args) -> ErrorCode:
 			for index, (source_path, target_path) in enumerate(itertools.product(source_paths, target_paths)):
 				step_args['source_paths'] = [ source_path ]
 				step_args['target_path'] = target_path
-				step_args['output_path'] = job_args.get('output_pattern').format(index = index)
+
+				try:
+					step_args['output_path'] = job_args.get('output_pattern').format(index = index, source_name = get_file_name(source_path), target_name = get_file_name(target_path), target_extension = get_file_extension(target_path))
+				except KeyError:
+					return 1
+
 				if not job_manager.add_step(job_id, step_args):
 					return 1
 			if job_manager.submit_job(job_id) and job_runner.run_job(job_id, process_step):
@@ -304,7 +284,12 @@ def process_batch(args : Args) -> ErrorCode:
 		if not source_paths and target_paths:
 			for index, target_path in enumerate(target_paths):
 				step_args['target_path'] = target_path
-				step_args['output_path'] = job_args.get('output_pattern').format(index = index)
+
+				try:
+					step_args['output_path'] = job_args.get('output_pattern').format(index = index, target_name = get_file_name(target_path), target_extension = get_file_extension(target_path))
+				except KeyError:
+					return 1
+
 				if not job_manager.add_step(job_id, step_args):
 					return 1
 			if job_manager.submit_job(job_id) and job_runner.run_job(job_id, process_step):
@@ -313,12 +298,11 @@ def process_batch(args : Args) -> ErrorCode:
 
 
 def process_step(job_id : str, step_index : int, step_args : Args) -> bool:
-	clear_reference_faces()
 	step_total = job_manager.count_step_total(job_id)
 	step_args.update(collect_job_args())
 	apply_args(step_args, state_manager.set_item)
 
-	logger.info(wording.get('processing_step').format(step_current = step_index + 1, step_total = step_total), __name__)
+	logger.info(translator.get('processing_step').format(step_current = step_index + 1, step_total = step_total), __name__)
 	if common_pre_check() and processors_pre_check():
 		error_code = conditional_process()
 		return error_code == 0
@@ -328,180 +312,25 @@ def process_step(job_id : str, step_index : int, step_args : Args) -> bool:
 def conditional_process() -> ErrorCode:
 	start_time = time()
 
+	if state_manager.get_item('workflow_mode') == 'auto':
+		state_manager.set_item('workflow_mode', detect_workflow_mode())
+
 	for processor_module in get_processors_modules(state_manager.get_item('processors')):
 		if not processor_module.pre_process('output'):
 			return 2
 
-	conditional_append_reference_faces()
+	if state_manager.get_item('workflow_mode') == 'image-to-image':
+		return image_to_image.process(start_time)
+	if state_manager.get_item('workflow_mode') == 'image-to-video':
+		return image_to_video.process(start_time)
 
-	if is_image(state_manager.get_item('target_path')):
-		return process_image(start_time)
+	return 0
+
+
+def detect_workflow_mode() -> WorkflowMode:
 	if is_video(state_manager.get_item('target_path')):
-		return process_video(start_time)
+		return 'image-to-video'
 
-	return 0
+	return 'image-to-image'
 
 
-def conditional_append_reference_faces() -> None:
-	if 'reference' in state_manager.get_item('face_selector_mode') and not get_reference_faces():
-		source_frames = read_static_images(state_manager.get_item('source_paths'))
-		source_faces = get_many_faces(source_frames)
-		source_face = get_average_face(source_faces)
-		if is_video(state_manager.get_item('target_path')):
-			reference_frame = read_video_frame(state_manager.get_item('target_path'), state_manager.get_item('reference_frame_number'))
-		else:
-			reference_frame = read_image(state_manager.get_item('target_path'))
-		reference_faces = sort_and_filter_faces(get_many_faces([ reference_frame ]))
-		reference_face = get_one_face(reference_faces, state_manager.get_item('reference_face_position'))
-		append_reference_face('origin', reference_face)
-
-		if source_face and reference_face:
-			for processor_module in get_processors_modules(state_manager.get_item('processors')):
-				abstract_reference_frame = processor_module.get_reference_frame(source_face, reference_face, reference_frame)
-				if numpy.any(abstract_reference_frame):
-					abstract_reference_faces = sort_and_filter_faces(get_many_faces([ abstract_reference_frame ]))
-					abstract_reference_face = get_one_face(abstract_reference_faces, state_manager.get_item('reference_face_position'))
-					append_reference_face(processor_module.__name__, abstract_reference_face)
-
-
-def process_image(start_time : float) -> ErrorCode:
-	if analyse_image(state_manager.get_item('target_path')):
-		return 3
-
-	logger.debug(wording.get('clearing_temp'), __name__)
-	clear_temp_directory(state_manager.get_item('target_path'))
-	logger.debug(wording.get('creating_temp'), __name__)
-	create_temp_directory(state_manager.get_item('target_path'))
-
-	process_manager.start()
-	temp_image_resolution = pack_resolution(restrict_image_resolution(state_manager.get_item('target_path'), unpack_resolution(state_manager.get_item('output_image_resolution'))))
-	logger.info(wording.get('copying_image').format(resolution = temp_image_resolution), __name__)
-	if copy_image(state_manager.get_item('target_path'), temp_image_resolution):
-		logger.debug(wording.get('copying_image_succeed'), __name__)
-	else:
-		logger.error(wording.get('copying_image_failed'), __name__)
-		process_manager.end()
-		return 1
-
-	temp_file_path = get_temp_file_path(state_manager.get_item('target_path'))
-	for processor_module in get_processors_modules(state_manager.get_item('processors')):
-		logger.info(wording.get('processing'), processor_module.__name__)
-		processor_module.process_image(state_manager.get_item('source_paths'), temp_file_path, temp_file_path)
-		processor_module.post_process()
-	if is_process_stopping():
-		process_manager.end()
-		return 4
-
-	logger.info(wording.get('finalizing_image').format(resolution = state_manager.get_item('output_image_resolution')), __name__)
-	if finalize_image(state_manager.get_item('target_path'), state_manager.get_item('output_path'), state_manager.get_item('output_image_resolution')):
-		logger.debug(wording.get('finalizing_image_succeed'), __name__)
-	else:
-		logger.warn(wording.get('finalizing_image_skipped'), __name__)
-
-	logger.debug(wording.get('clearing_temp'), __name__)
-	clear_temp_directory(state_manager.get_item('target_path'))
-
-	if is_image(state_manager.get_item('output_path')):
-		seconds = '{:.2f}'.format((time() - start_time) % 60)
-		logger.info(wording.get('processing_image_succeed').format(seconds = seconds), __name__)
-		conditional_log_statistics()
-	else:
-		logger.error(wording.get('processing_image_failed'), __name__)
-		process_manager.end()
-		return 1
-	process_manager.end()
-	return 0
-
-
-def process_video(start_time : float) -> ErrorCode:
-	trim_frame_start, trim_frame_end = restrict_trim_frame(state_manager.get_item('target_path'), state_manager.get_item('trim_frame_start'), state_manager.get_item('trim_frame_end'))
-	if analyse_video(state_manager.get_item('target_path'), trim_frame_start, trim_frame_end):
-		return 3
-
-	logger.debug(wording.get('clearing_temp'), __name__)
-	clear_temp_directory(state_manager.get_item('target_path'))
-	logger.debug(wording.get('creating_temp'), __name__)
-	create_temp_directory(state_manager.get_item('target_path'))
-
-	process_manager.start()
-	temp_video_resolution = pack_resolution(restrict_video_resolution(state_manager.get_item('target_path'), unpack_resolution(state_manager.get_item('output_video_resolution'))))
-	temp_video_fps = restrict_video_fps(state_manager.get_item('target_path'), state_manager.get_item('output_video_fps'))
-	logger.info(wording.get('extracting_frames').format(resolution = temp_video_resolution, fps = temp_video_fps), __name__)
-	if extract_frames(state_manager.get_item('target_path'), temp_video_resolution, temp_video_fps, trim_frame_start, trim_frame_end):
-		logger.debug(wording.get('extracting_frames_succeed'), __name__)
-	else:
-		if is_process_stopping():
-			process_manager.end()
-			return 4
-		logger.error(wording.get('extracting_frames_failed'), __name__)
-		process_manager.end()
-		return 1
-
-	temp_frame_paths = resolve_temp_frame_paths(state_manager.get_item('target_path'))
-	if temp_frame_paths:
-		for processor_module in get_processors_modules(state_manager.get_item('processors')):
-			logger.info(wording.get('processing'), processor_module.__name__)
-			processor_module.process_video(state_manager.get_item('source_paths'), temp_frame_paths)
-			processor_module.post_process()
-		if is_process_stopping():
-			return 4
-	else:
-		logger.error(wording.get('temp_frames_not_found'), __name__)
-		process_manager.end()
-		return 1
-
-	logger.info(wording.get('merging_video').format(resolution = state_manager.get_item('output_video_resolution'), fps = state_manager.get_item('output_video_fps')), __name__)
-	if merge_video(state_manager.get_item('target_path'), temp_video_fps, state_manager.get_item('output_video_resolution'), state_manager.get_item('output_video_fps'), trim_frame_start, trim_frame_end):
-		logger.debug(wording.get('merging_video_succeed'), __name__)
-	else:
-		if is_process_stopping():
-			process_manager.end()
-			return 4
-		logger.error(wording.get('merging_video_failed'), __name__)
-		process_manager.end()
-		return 1
-
-	if state_manager.get_item('output_audio_volume') == 0:
-		logger.info(wording.get('skipping_audio'), __name__)
-		move_temp_file(state_manager.get_item('target_path'), state_manager.get_item('output_path'))
-	else:
-		source_audio_path = get_first(filter_audio_paths(state_manager.get_item('source_paths')))
-		if source_audio_path:
-			if replace_audio(state_manager.get_item('target_path'), source_audio_path, state_manager.get_item('output_path')):
-				logger.debug(wording.get('replacing_audio_succeed'), __name__)
-			else:
-				if is_process_stopping():
-					process_manager.end()
-					return 4
-				logger.warn(wording.get('replacing_audio_skipped'), __name__)
-				move_temp_file(state_manager.get_item('target_path'), state_manager.get_item('output_path'))
-		else:
-			if restore_audio(state_manager.get_item('target_path'), state_manager.get_item('output_path'), trim_frame_start, trim_frame_end):
-				logger.debug(wording.get('restoring_audio_succeed'), __name__)
-			else:
-				if is_process_stopping():
-					process_manager.end()
-					return 4
-				logger.warn(wording.get('restoring_audio_skipped'), __name__)
-				move_temp_file(state_manager.get_item('target_path'), state_manager.get_item('output_path'))
-
-	logger.debug(wording.get('clearing_temp'), __name__)
-	clear_temp_directory(state_manager.get_item('target_path'))
-
-	if is_video(state_manager.get_item('output_path')):
-		seconds = '{:.2f}'.format((time() - start_time))
-		logger.info(wording.get('processing_video_succeed').format(seconds = seconds), __name__)
-		conditional_log_statistics()
-	else:
-		logger.error(wording.get('processing_video_failed'), __name__)
-		process_manager.end()
-		return 1
-	process_manager.end()
-	return 0
-
-
-def is_process_stopping() -> bool:
-	if process_manager.is_stopping():
-		process_manager.end()
-		logger.info(wording.get('processing_stopped'), __name__)
-	return process_manager.is_pending()

facefusion/curl_builder.py

@@ -1,27 +1,32 @@
 import itertools
 import shutil
+from typing import List
 
 from facefusion import metadata
-from facefusion.types import Commands
+from facefusion.types import Command
 
 
-def run(commands : Commands) -> Commands:
+def run(commands : List[Command]) -> List[Command]:
 	user_agent = metadata.get('name') + '/' + metadata.get('version')
 
-	return [ shutil.which('curl'), '--user-agent', user_agent, '--insecure', '--location', '--silent' ] + commands
+	return [ shutil.which('curl'), '--user-agent', user_agent, '--location', '--silent', '--ssl-no-revoke' ] + commands
 
 
-def chain(*commands : Commands) -> Commands:
+def chain(*commands : List[Command]) -> List[Command]:
 	return list(itertools.chain(*commands))
 
 
-def head(url : str) -> Commands:
+def ping(url : str) -> List[Command]:
 	return [ '-I', url ]
 
 
-def download(url : str, download_file_path : str) -> Commands:
+def download(url : str, download_file_path : str) -> List[Command]:
 	return [ '--create-dirs', '--continue-at', '-', '--output', download_file_path, url ]
 
 
-def set_timeout(timeout : int) -> Commands:
+def set_timeout(timeout : int) -> List[Command]:
 	return [ '--connect-timeout', str(timeout) ]
+
+
+def set_retry(retry : int) -> List[Command]:
+	return [ '--retry', str(retry) ]

facefusion/download.py

@@ -7,13 +7,13 @@ from urllib.parse import urlparse
 from tqdm import tqdm
 
 import facefusion.choices
-from facefusion import curl_builder, logger, process_manager, state_manager, wording
+from facefusion import curl_builder, logger, process_manager, state_manager, translator
 from facefusion.filesystem import get_file_name, get_file_size, is_file, remove_file
 from facefusion.hash_helper import validate_hash
-from facefusion.types import Commands, DownloadProvider, DownloadSet
+from facefusion.types import Command, DownloadProvider, DownloadSet
 
 
-def open_curl(commands : Commands) -> subprocess.Popen[bytes]:
+def open_curl(commands : List[Command]) -> subprocess.Popen[bytes]:
 	commands = curl_builder.run(commands)
 	return subprocess.Popen(commands, stdin = subprocess.PIPE, stdout = subprocess.PIPE)
 
@@ -26,11 +26,13 @@ def conditional_download(download_directory_path : str, urls : List[str]) -> Non
 		download_size = get_static_download_size(url)
 
 		if initial_size < download_size:
-			with tqdm(total = download_size, initial = initial_size, desc = wording.get('downloading'), unit = 'B', unit_scale = True, unit_divisor = 1024, ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
+			with tqdm(total = download_size, initial = initial_size, desc = translator.get('downloading'), unit = 'B', unit_scale = True, unit_divisor = 1024, ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
 				commands = curl_builder.chain(
 					curl_builder.download(url, download_file_path),
-					curl_builder.set_timeout(10)
+					curl_builder.set_timeout(5),
+					curl_builder.set_retry(5)
 				)
+
 				open_curl(commands)
 				current_size = initial_size
 				progress.set_postfix(download_providers = state_manager.get_item('download_providers'), file_name = download_file_name)
@@ -41,12 +43,13 @@ def conditional_download(download_directory_path : str, urls : List[str]) -> Non
 						progress.update(current_size - progress.n)
 
 
-@lru_cache(maxsize = None)
+@lru_cache(maxsize = 64)
 def get_static_download_size(url : str) -> int:
 	commands = curl_builder.chain(
-		curl_builder.head(url),
+		curl_builder.ping(url),
 		curl_builder.set_timeout(5)
 	)
+
 	process = open_curl(commands)
 	lines = reversed(process.stdout.readlines())
 
@@ -59,12 +62,13 @@ def get_static_download_size(url : str) -> int:
 	return 0
 
 
-@lru_cache(maxsize = None)
+@lru_cache(maxsize = 64)
 def ping_static_url(url : str) -> bool:
 	commands = curl_builder.chain(
-		curl_builder.head(url),
+		curl_builder.ping(url),
 		curl_builder.set_timeout(5)
 	)
+
 	process = open_curl(commands)
 	process.communicate()
 	return process.returncode == 0
@@ -87,10 +91,10 @@ def conditional_download_hashes(hash_set : DownloadSet) -> bool:
 
 	for valid_hash_path in valid_hash_paths:
 		valid_hash_file_name = get_file_name(valid_hash_path)
-		logger.debug(wording.get('validating_hash_succeed').format(hash_file_name = valid_hash_file_name), __name__)
+		logger.debug(translator.get('validating_hash_succeeded').format(hash_file_name = valid_hash_file_name), __name__)
 	for invalid_hash_path in invalid_hash_paths:
 		invalid_hash_file_name = get_file_name(invalid_hash_path)
-		logger.error(wording.get('validating_hash_failed').format(hash_file_name = invalid_hash_file_name), __name__)
+		logger.error(translator.get('validating_hash_failed').format(hash_file_name = invalid_hash_file_name), __name__)
 
 	if not invalid_hash_paths:
 		process_manager.end()
@@ -114,13 +118,13 @@ def conditional_download_sources(source_set : DownloadSet) -> bool:
 
 	for valid_source_path in valid_source_paths:
 		valid_source_file_name = get_file_name(valid_source_path)
-		logger.debug(wording.get('validating_source_succeed').format(source_file_name = valid_source_file_name), __name__)
+		logger.debug(translator.get('validating_source_succeeded').format(source_file_name = valid_source_file_name), __name__)
 	for invalid_source_path in invalid_source_paths:
 		invalid_source_file_name = get_file_name(invalid_source_path)
-		logger.error(wording.get('validating_source_failed').format(source_file_name = invalid_source_file_name), __name__)
+		logger.error(translator.get('validating_source_failed').format(source_file_name = invalid_source_file_name), __name__)
 
 		if remove_file(invalid_source_path):
-			logger.error(wording.get('deleting_corrupt_source').format(source_file_name = invalid_source_file_name), __name__)
+			logger.error(translator.get('deleting_corrupt_source').format(source_file_name = invalid_source_file_name), __name__)
 
 	if not invalid_source_paths:
 		process_manager.end()

facefusion/execution.py

@@ -1,15 +1,17 @@
+import os
 import shutil
 import subprocess
 import xml.etree.ElementTree as ElementTree
 from functools import lru_cache
 from typing import List, Optional
 
-from onnxruntime import get_available_providers, set_default_logger_severity
+import onnxruntime
 
 import facefusion.choices
-from facefusion.types import ExecutionDevice, ExecutionProvider, InferenceSessionProvider, ValueAndUnit
+from facefusion.filesystem import create_directory, is_directory
+from facefusion.types import ExecutionDevice, ExecutionProvider, InferenceOptionSet, InferenceProvider, ValueAndUnit
 
-set_default_logger_severity(3)
+onnxruntime.set_default_logger_severity(3)
 
 
 def has_execution_provider(execution_provider : ExecutionProvider) -> bool:
@@ -17,7 +19,7 @@ def has_execution_provider(execution_provider : ExecutionProvider) -> bool:
 
 
 def get_available_execution_providers() -> List[ExecutionProvider]:
-	inference_session_providers = get_available_providers()
+	inference_session_providers = onnxruntime.get_available_providers()
 	available_execution_providers : List[ExecutionProvider] = []
 
 	for execution_provider, execution_provider_value in facefusion.choices.execution_provider_set.items():
@@ -28,48 +30,86 @@ def get_available_execution_providers() -> List[ExecutionProvider]:
 	return available_execution_providers
 
 
-def create_inference_session_providers(execution_device_id : str, execution_providers : List[ExecutionProvider]) -> List[InferenceSessionProvider]:
-	inference_session_providers : List[InferenceSessionProvider] = []
+def create_inference_providers(execution_device_id : int, execution_providers : List[ExecutionProvider]) -> List[InferenceProvider]:
+	inference_providers : List[InferenceProvider] = []
+	cache_path = resolve_cache_path()
 
 	for execution_provider in execution_providers:
 		if execution_provider == 'cuda':
-			inference_session_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
 			{
 				'device_id': execution_device_id,
 				'cudnn_conv_algo_search': resolve_cudnn_conv_algo_search()
 			}))
+
 		if execution_provider == 'tensorrt':
-			inference_session_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
+			inference_option_set : InferenceOptionSet =\
 			{
-				'device_id': execution_device_id,
-				'trt_engine_cache_enable': True,
-				'trt_engine_cache_path': '.caches',
-				'trt_timing_cache_enable': True,
-				'trt_timing_cache_path': '.caches',
-				'trt_builder_optimization_level': 5
-			}))
+				'device_id': execution_device_id
+			}
+			if is_directory(cache_path) or create_directory(cache_path):
+				inference_option_set.update(
+				{
+					'trt_engine_cache_enable': True,
+					'trt_engine_cache_path': cache_path,
+					'trt_timing_cache_enable': True,
+					'trt_timing_cache_path': cache_path,
+					'trt_builder_optimization_level': 4
+				})
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider), inference_option_set))
+
 		if execution_provider in [ 'directml', 'rocm' ]:
-			inference_session_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
 			{
 				'device_id': execution_device_id
 			}))
+
+		if execution_provider == 'migraphx':
+			inference_option_set =\
+			{
+				'device_id': execution_device_id
+			}
+			if is_directory(cache_path) or create_directory(cache_path):
+				inference_option_set.update(
+				{
+					'migraphx_model_cache_dir': cache_path
+				})
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider), inference_option_set))
+
+		if execution_provider == 'coreml':
+			inference_option_set =\
+			{
+				'SpecializationStrategy': 'FastPrediction'
+			}
+			if is_directory(cache_path) or create_directory(cache_path):
+				inference_option_set.update(
+				{
+					'ModelCacheDirectory': cache_path
+				})
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider), inference_option_set))
+
 		if execution_provider == 'openvino':
-			inference_session_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
 			{
 				'device_type': resolve_openvino_device_type(execution_device_id),
 				'precision': 'FP32'
 			}))
-		if execution_provider == 'coreml':
-			inference_session_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
+
+		if execution_provider == 'qnn':
+			inference_providers.append((facefusion.choices.execution_provider_set.get(execution_provider),
 			{
-				'SpecializationStrategy': 'FastPrediction',
-				'ModelCacheDirectory': '.caches'
+				'device_id': execution_device_id,
+				'backend_type': 'htp'
 			}))
 
 	if 'cpu' in execution_providers:
-		inference_session_providers.append(facefusion.choices.execution_provider_set.get('cpu'))
+		inference_providers.append(facefusion.choices.execution_provider_set.get('cpu'))
 
-	return inference_session_providers
+	return inference_providers
+
+
+def resolve_cache_path() -> str:
+	return os.path.join('.caches', onnxruntime.get_version_string())
 
 
 def resolve_cudnn_conv_algo_search() -> str:
@@ -83,12 +123,10 @@ def resolve_cudnn_conv_algo_search() -> str:
 	return 'EXHAUSTIVE'
 
 
-def resolve_openvino_device_type(execution_device_id : str) -> str:
-	if execution_device_id == '0':
+def resolve_openvino_device_type(execution_device_id : int) -> str:
+	if execution_device_id == 0:
 		return 'GPU'
-	if execution_device_id == '∞':
-		return 'MULTI:GPU'
-	return 'GPU.' + execution_device_id
+	return 'GPU.' + str(execution_device_id)
 
 
 def run_nvidia_smi() -> subprocess.Popen[bytes]:
@@ -96,7 +134,7 @@ def run_nvidia_smi() -> subprocess.Popen[bytes]:
 	return subprocess.Popen(commands, stdout = subprocess.PIPE)
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def detect_static_execution_devices() -> List[ExecutionDevice]:
 	return detect_execution_devices()
 

facefusion/exit_helper.py

@@ -1,21 +1,34 @@
+import os
 import signal
 import sys
 from time import sleep
+from types import FrameType
 
 from facefusion import process_manager, state_manager
 from facefusion.temp_helper import clear_temp_directory
 from facefusion.types import ErrorCode
 
 
+def fatal_exit(error_code : ErrorCode) -> None:
+	os._exit(error_code)
+
+
 def hard_exit(error_code : ErrorCode) -> None:
-	signal.signal(signal.SIGINT, signal.SIG_IGN)
 	sys.exit(error_code)
 
 
+def signal_exit(signum : int, frame : FrameType) -> None:
+	graceful_exit(0)
+
+
 def graceful_exit(error_code : ErrorCode) -> None:
+	signal.signal(signal.SIGINT, signal.SIG_IGN)
 	process_manager.stop()
+
 	while process_manager.is_processing():
 		sleep(0.5)
+
 	if state_manager.get_item('target_path'):
 		clear_temp_directory(state_manager.get_item('target_path'))
+
 	hard_exit(error_code)

facefusion/face_analyser.py

@@ -1,124 +0,0 @@
-from typing import List, Optional
-
-import numpy
-
-from facefusion import state_manager
-from facefusion.common_helper import get_first
-from facefusion.face_classifier import classify_face
-from facefusion.face_detector import detect_faces, detect_rotated_faces
-from facefusion.face_helper import apply_nms, convert_to_face_landmark_5, estimate_face_angle, get_nms_threshold
-from facefusion.face_landmarker import detect_face_landmark, estimate_face_landmark_68_5
-from facefusion.face_recognizer import calc_embedding
-from facefusion.face_store import get_static_faces, set_static_faces
-from facefusion.types import BoundingBox, Face, FaceLandmark5, FaceLandmarkSet, FaceScoreSet, Score, VisionFrame
-
-
-def create_faces(vision_frame : VisionFrame, bounding_boxes : List[BoundingBox], face_scores : List[Score], face_landmarks_5 : List[FaceLandmark5]) -> List[Face]:
-	faces = []
-	nms_threshold = get_nms_threshold(state_manager.get_item('face_detector_model'), state_manager.get_item('face_detector_angles'))
-	keep_indices = apply_nms(bounding_boxes, face_scores, state_manager.get_item('face_detector_score'), nms_threshold)
-
-	for index in keep_indices:
-		bounding_box = bounding_boxes[index]
-		face_score = face_scores[index]
-		face_landmark_5 = face_landmarks_5[index]
-		face_landmark_5_68 = face_landmark_5
-		face_landmark_68_5 = estimate_face_landmark_68_5(face_landmark_5_68)
-		face_landmark_68 = face_landmark_68_5
-		face_landmark_score_68 = 0.0
-		face_angle = estimate_face_angle(face_landmark_68_5)
-
-		if state_manager.get_item('face_landmarker_score') > 0:
-			face_landmark_68, face_landmark_score_68 = detect_face_landmark(vision_frame, bounding_box, face_angle)
-		if face_landmark_score_68 > state_manager.get_item('face_landmarker_score'):
-			face_landmark_5_68 = convert_to_face_landmark_5(face_landmark_68)
-
-		face_landmark_set : FaceLandmarkSet =\
-		{
-			'5': face_landmark_5,
-			'5/68': face_landmark_5_68,
-			'68': face_landmark_68,
-			'68/5': face_landmark_68_5
-		}
-		face_score_set : FaceScoreSet =\
-		{
-			'detector': face_score,
-			'landmarker': face_landmark_score_68
-		}
-		embedding, normed_embedding = calc_embedding(vision_frame, face_landmark_set.get('5/68'))
-		gender, age, race = classify_face(vision_frame, face_landmark_set.get('5/68'))
-		faces.append(Face(
-			bounding_box = bounding_box,
-			score_set = face_score_set,
-			landmark_set = face_landmark_set,
-			angle = face_angle,
-			embedding = embedding,
-			normed_embedding = normed_embedding,
-			gender = gender,
-			age = age,
-			race = race
-		))
-	return faces
-
-
-def get_one_face(faces : List[Face], position : int = 0) -> Optional[Face]:
-	if faces:
-		position = min(position, len(faces) - 1)
-		return faces[position]
-	return None
-
-
-def get_average_face(faces : List[Face]) -> Optional[Face]:
-	embeddings = []
-	normed_embeddings = []
-
-	if faces:
-		first_face = get_first(faces)
-
-		for face in faces:
-			embeddings.append(face.embedding)
-			normed_embeddings.append(face.normed_embedding)
-
-		return Face(
-			bounding_box = first_face.bounding_box,
-			score_set = first_face.score_set,
-			landmark_set = first_face.landmark_set,
-			angle = first_face.angle,
-			embedding = numpy.mean(embeddings, axis = 0),
-			normed_embedding = numpy.mean(normed_embeddings, axis = 0),
-			gender = first_face.gender,
-			age = first_face.age,
-			race = first_face.race
-		)
-	return None
-
-
-def get_many_faces(vision_frames : List[VisionFrame]) -> List[Face]:
-	many_faces : List[Face] = []
-
-	for vision_frame in vision_frames:
-		if numpy.any(vision_frame):
-			static_faces = get_static_faces(vision_frame)
-			if static_faces:
-				many_faces.extend(static_faces)
-			else:
-				all_bounding_boxes = []
-				all_face_scores = []
-				all_face_landmarks_5 = []
-
-				for face_detector_angle in state_manager.get_item('face_detector_angles'):
-					if face_detector_angle == 0:
-						bounding_boxes, face_scores, face_landmarks_5 = detect_faces(vision_frame)
-					else:
-						bounding_boxes, face_scores, face_landmarks_5 = detect_rotated_faces(vision_frame, face_detector_angle)
-					all_bounding_boxes.extend(bounding_boxes)
-					all_face_scores.extend(face_scores)
-					all_face_landmarks_5.extend(face_landmarks_5)
-
-				if all_bounding_boxes and all_face_scores and all_face_landmarks_5 and state_manager.get_item('face_detector_score') > 0:
-					faces = create_faces(vision_frame, all_bounding_boxes, all_face_scores, all_face_landmarks_5)
-
-					if faces:
-						many_faces.extend(faces)
-						set_static_faces(vision_frame, faces)
-	return many_faces

facefusion/face_classifier.py

@@ -11,12 +11,18 @@ from facefusion.thread_helper import conditional_thread_semaphore
 from facefusion.types import Age, DownloadScope, FaceLandmark5, Gender, InferencePool, ModelOptions, ModelSet, Race, VisionFrame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'fairface':
 		{
+			'__metadata__':
+			{
+				'vendor': 'dchen236',
+				'license': 'CC-BY-4.0',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_classifier':

facefusion/face_creator.py

@@ -0,0 +1,209 @@
+from typing import List, Optional
+
+import numpy
+
+from facefusion import face_store, state_manager
+from facefusion.common_helper import get_first, get_middle
+from facefusion.face_classifier import classify_face
+from facefusion.face_detector import detect_faces, detect_faces_by_angle
+from facefusion.face_helper import apply_nms, average_points, convert_to_face_landmark_5, estimate_face_angle, get_nms_threshold
+from facefusion.face_landmarker import detect_face_landmark, estimate_face_landmark_68_5
+from facefusion.face_recognizer import calculate_face_embedding
+from facefusion.types import BoundingBox, Face, FaceLandmark5, FaceLandmarkSet, FaceScoreSet, Score, VisionFrame
+
+
+def create_faces(vision_frame : VisionFrame, bounding_boxes : List[BoundingBox], face_scores : List[Score], face_landmarks_5 : List[FaceLandmark5]) -> List[Face]:
+	faces = []
+	nms_threshold = get_nms_threshold(state_manager.get_item('face_detector_model'), state_manager.get_item('face_detector_angles'))
+	keep_indices = apply_nms(bounding_boxes, face_scores, state_manager.get_item('face_detector_score'), nms_threshold)
+
+	for index in keep_indices:
+		bounding_box = bounding_boxes[index]
+		face_score = face_scores[index]
+		face_landmark_5 = face_landmarks_5[index]
+		face_landmark_5_68 = face_landmark_5
+		face_landmark_68_5 = estimate_face_landmark_68_5(face_landmark_5_68)
+		face_landmark_68 = face_landmark_68_5
+		face_landmark_score_68 = 0.0
+		face_angle = estimate_face_angle(face_landmark_68_5)
+
+		if state_manager.get_item('face_landmarker_score') > 0:
+			face_landmark_68, face_landmark_score_68 = detect_face_landmark(vision_frame, bounding_box, face_angle)
+		if face_landmark_score_68 > state_manager.get_item('face_landmarker_score'):
+			face_landmark_5_68 = convert_to_face_landmark_5(face_landmark_68)
+
+		face_landmark_set : FaceLandmarkSet =\
+		{
+			'5': face_landmark_5,
+			'5/68': face_landmark_5_68,
+			'68': face_landmark_68,
+			'68/5': face_landmark_68_5
+		}
+		face_score_set : FaceScoreSet =\
+		{
+			'detector': face_score,
+			'landmarker': face_landmark_score_68
+		}
+		face_embedding, face_embedding_norm = calculate_face_embedding(vision_frame, face_landmark_set.get('5/68'))
+		gender, age, race = classify_face(vision_frame, face_landmark_set.get('5/68'))
+
+		faces.append(Face(
+			origin = 'detect',
+			bounding_box = bounding_box,
+			score_set = face_score_set,
+			landmark_set = face_landmark_set,
+			angle = face_angle,
+			embedding = face_embedding,
+			embedding_norm = face_embedding_norm,
+			gender = gender,
+			age = age,
+			race = race
+		))
+	return faces
+
+
+def get_one_face(faces : List[Face], position : int = 0) -> Optional[Face]:
+	if faces:
+		position = min(position, len(faces) - 1)
+		return faces[position]
+	return None
+
+
+def get_many_faces(vision_frames : List[VisionFrame]) -> List[Face]:
+	many_faces : List[Face] = []
+
+	for vision_frame in vision_frames:
+		if numpy.any(vision_frame):
+			all_bounding_boxes = []
+			all_face_scores = []
+			all_face_landmarks_5 = []
+
+			for face_detector_angle in state_manager.get_item('face_detector_angles'):
+				if face_detector_angle == 0:
+					bounding_boxes, face_scores, face_landmarks_5 = detect_faces(vision_frame)
+				else:
+					bounding_boxes, face_scores, face_landmarks_5 = detect_faces_by_angle(vision_frame, face_detector_angle)
+				all_bounding_boxes.extend(bounding_boxes)
+				all_face_scores.extend(face_scores)
+				all_face_landmarks_5.extend(face_landmarks_5)
+
+			if all_bounding_boxes and all_face_scores and all_face_landmarks_5 and state_manager.get_item('face_detector_score') > 0:
+				faces = create_faces(vision_frame, all_bounding_boxes, all_face_scores, all_face_landmarks_5)
+
+				if faces:
+					many_faces.extend(faces)
+
+	return many_faces
+
+
+def get_static_faces(vision_frames : List[VisionFrame]) -> List[Face]:
+	many_faces : List[Face] = []
+
+	for vision_frame in vision_frames:
+		faces = face_store.get_faces(vision_frame)
+
+		if not faces:
+			with face_store.resolve_lock(vision_frame):
+				faces = face_store.get_faces(vision_frame)
+
+				if not faces:
+					faces = get_many_faces([ vision_frame ])
+
+					if faces:
+						face_store.set_faces(vision_frame, faces)
+
+		many_faces.extend(faces)
+
+	return many_faces
+
+
+def refill_faces(faces : List[Optional[Face]]) -> List[Face]:
+	fill_faces = []
+	anchor_index_previous = -1
+
+	for index, face in enumerate(faces):
+		if face:
+			for gap_index in range(anchor_index_previous + 1, index):
+				average_factor = (gap_index - anchor_index_previous) / (index - anchor_index_previous)
+				average_face = average_face_geometry([faces[anchor_index_previous], face], average_factor)
+				fill_faces.append(average_face)
+
+			fill_faces.append(face)
+			anchor_index_previous = index
+
+	return fill_faces
+
+
+def average_face_geometry(faces : List[Face], average_factor : float) -> Face:
+	face_first = get_first(faces)
+	face_middle = get_middle(faces)
+	face_anchor = face_middle
+
+	if average_factor < 0.5:
+		face_anchor = face_first
+
+	landmark_set : FaceLandmarkSet =\
+	{
+		'5': average_points(face_first.landmark_set.get('5'), face_middle.landmark_set.get('5'), average_factor),
+		'5/68': average_points(face_first.landmark_set.get('5/68'), face_middle.landmark_set.get('5/68'), average_factor),
+		'68': average_points(face_first.landmark_set.get('68'), face_middle.landmark_set.get('68'), average_factor),
+		'68/5': average_points(face_first.landmark_set.get('68/5'), face_middle.landmark_set.get('68/5'), average_factor)
+	}
+
+	return Face(
+		origin = 'refill',
+		bounding_box = average_points(face_first.bounding_box, face_middle.bounding_box, average_factor),
+		score_set = face_anchor.score_set,
+		landmark_set = landmark_set,
+		angle = estimate_face_angle(landmark_set.get('68/5')),
+		embedding = face_anchor.embedding,
+		embedding_norm = face_anchor.embedding_norm,
+		gender = face_anchor.gender,
+		age = face_anchor.age,
+		race = face_anchor.race
+	)
+
+
+def average_face_identity(faces : List[Face]) -> Optional[Face]:
+	face_embeddings = []
+	face_embeddings_norm = []
+
+	if faces:
+		first_face = get_first(faces)
+
+		for face in faces:
+			face_embeddings.append(face.embedding)
+			face_embeddings_norm.append(face.embedding_norm)
+
+		return Face(
+			origin = first_face.origin,
+			bounding_box = first_face.bounding_box,
+			score_set = first_face.score_set,
+			landmark_set = first_face.landmark_set,
+			angle = first_face.angle,
+			embedding = numpy.mean(face_embeddings, axis = 0),
+			embedding_norm = numpy.mean(face_embeddings_norm, axis = 0),
+			gender = first_face.gender,
+			age = first_face.age,
+			race = first_face.race
+		)
+	return None
+
+
+def scale_face(target_face : Face, target_vision_frame : VisionFrame, temp_vision_frame : VisionFrame) -> Face:
+	scale_x = temp_vision_frame.shape[1] / target_vision_frame.shape[1]
+	scale_y = temp_vision_frame.shape[0] / target_vision_frame.shape[0]
+
+	bounding_box = target_face.bounding_box * [ scale_x, scale_y, scale_x, scale_y ]
+	landmark_set =\
+	{
+		'5': target_face.landmark_set.get('5') * numpy.array([ scale_x, scale_y ]),
+		'5/68': target_face.landmark_set.get('5/68') * numpy.array([ scale_x, scale_y ]),
+		'68': target_face.landmark_set.get('68') * numpy.array([ scale_x, scale_y ]),
+		'68/5': target_face.landmark_set.get('68/5') * numpy.array([ scale_x, scale_y ])
+	}
+
+	return target_face._replace(
+		bounding_box = bounding_box,
+		landmark_set = landmark_set
+	)

facefusion/face_detector.py

@@ -6,19 +6,25 @@ import numpy
 
 from facefusion import inference_manager, state_manager
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.face_helper import create_rotated_matrix_and_size, create_static_anchors, distance_to_bounding_box, distance_to_face_landmark_5, normalize_bounding_box, transform_bounding_box, transform_points
+from facefusion.face_helper import create_rotation_matrix_and_size, create_static_anchors, distance_to_bounding_box, distance_to_face_landmark_5, normalize_bounding_box, transform_bounding_box, transform_points
 from facefusion.filesystem import resolve_relative_path
 from facefusion.thread_helper import thread_semaphore
-from facefusion.types import Angle, BoundingBox, Detection, DownloadScope, DownloadSet, FaceLandmark5, InferencePool, ModelSet, Score, VisionFrame
+from facefusion.types import Angle, BoundingBox, Detection, DownloadScope, DownloadSet, FaceLandmark5, InferencePool, Margin, ModelSet, Score, VisionFrame
 from facefusion.vision import restrict_frame, unpack_resolution
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'retinaface':
 		{
+			'__metadata__':
+			{
+				'vendor': 'InsightFace',
+				'license': 'Non-Commercial',
+				'year': 2020
+			},
 			'hashes':
 			{
 				'retinaface':
@@ -38,6 +44,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'scrfd':
 		{
+			'__metadata__':
+			{
+				'vendor': 'InsightFace',
+				'license': 'Non-Commercial',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'scrfd':
@@ -57,6 +69,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'yolo_face':
 		{
+			'__metadata__':
+			{
+				'vendor': 'derronqi',
+				'license': 'GPL-3.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'yolo_face':
@@ -73,6 +91,31 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/yoloface_8n.onnx')
 				}
 			}
+		},
+		'yunet':
+		{
+			'__metadata__':
+			{
+				'vendor': 'OpenCV',
+				'license': 'MIT',
+				'year': 2023
+			},
+			'hashes':
+			{
+				'yunet':
+				{
+					'url': resolve_download_url('models-3.4.0', 'yunet_2023_mar.hash'),
+					'path': resolve_relative_path('../.assets/models/yunet_2023_mar.hash')
+				}
+			},
+			'sources':
+			{
+				'yunet':
+				{
+					'url': resolve_download_url('models-3.4.0', 'yunet_2023_mar.onnx'),
+					'path': resolve_relative_path('../.assets/models/yunet_2023_mar.onnx')
+				}
+			}
 		}
 	}
 
@@ -94,7 +137,7 @@ def collect_model_downloads() -> Tuple[DownloadSet, DownloadSet]:
 	model_hash_set = {}
 	model_source_set = {}
 
-	for face_detector_model in [ 'retinaface', 'scrfd', 'yolo_face' ]:
+	for face_detector_model in [ 'retinaface', 'scrfd', 'yolo_face', 'yunet' ]:
 		if state_manager.get_item('face_detector_model') in [ 'many', face_detector_model ]:
 			model_hash_set[face_detector_model] = model_set.get(face_detector_model).get('hashes').get(face_detector_model)
 			model_source_set[face_detector_model] = model_set.get(face_detector_model).get('sources').get(face_detector_model)
@@ -109,39 +152,56 @@ def pre_check() -> bool:
 
 
 def detect_faces(vision_frame : VisionFrame) -> Tuple[List[BoundingBox], List[Score], List[FaceLandmark5]]:
+	margin_top, margin_right, margin_bottom, margin_left = prepare_margin(vision_frame)
+	margin_vision_frame = numpy.pad(vision_frame, ((margin_top, margin_bottom), (margin_left, margin_right), (0, 0)))
 	all_bounding_boxes : List[BoundingBox] = []
 	all_face_scores : List[Score] = []
 	all_face_landmarks_5 : List[FaceLandmark5] = []
 
 	if state_manager.get_item('face_detector_model') in [ 'many', 'retinaface' ]:
-		bounding_boxes, face_scores, face_landmarks_5 = detect_with_retinaface(vision_frame, state_manager.get_item('face_detector_size'))
+		bounding_boxes, face_scores, face_landmarks_5 = detect_with_retinaface(margin_vision_frame, state_manager.get_item('face_detector_size'))
 		all_bounding_boxes.extend(bounding_boxes)
 		all_face_scores.extend(face_scores)
 		all_face_landmarks_5.extend(face_landmarks_5)
 
 	if state_manager.get_item('face_detector_model') in [ 'many', 'scrfd' ]:
-		bounding_boxes, face_scores, face_landmarks_5 = detect_with_scrfd(vision_frame, state_manager.get_item('face_detector_size'))
+		bounding_boxes, face_scores, face_landmarks_5 = detect_with_scrfd(margin_vision_frame, state_manager.get_item('face_detector_size'))
 		all_bounding_boxes.extend(bounding_boxes)
 		all_face_scores.extend(face_scores)
 		all_face_landmarks_5.extend(face_landmarks_5)
 
 	if state_manager.get_item('face_detector_model') in [ 'many', 'yolo_face' ]:
-		bounding_boxes, face_scores, face_landmarks_5 = detect_with_yolo_face(vision_frame, state_manager.get_item('face_detector_size'))
+		bounding_boxes, face_scores, face_landmarks_5 = detect_with_yolo_face(margin_vision_frame, state_manager.get_item('face_detector_size'))
 		all_bounding_boxes.extend(bounding_boxes)
 		all_face_scores.extend(face_scores)
 		all_face_landmarks_5.extend(face_landmarks_5)
 
-	all_bounding_boxes = [ normalize_bounding_box(all_bounding_box) for all_bounding_box in all_bounding_boxes ]
+	if state_manager.get_item('face_detector_model') == 'yunet':
+		bounding_boxes, face_scores, face_landmarks_5 = detect_with_yunet(margin_vision_frame, state_manager.get_item('face_detector_size'))
+		all_bounding_boxes.extend(bounding_boxes)
+		all_face_scores.extend(face_scores)
+		all_face_landmarks_5.extend(face_landmarks_5)
+
+	all_bounding_boxes = [ normalize_bounding_box(all_bounding_box) - numpy.array([ margin_left, margin_top, margin_left, margin_top ]) for all_bounding_box in all_bounding_boxes ]
+	all_face_landmarks_5 = [ all_face_landmark_5 - numpy.array([ margin_left, margin_top ]) for all_face_landmark_5 in all_face_landmarks_5 ]
 	return all_bounding_boxes, all_face_scores, all_face_landmarks_5
 
 
-def detect_rotated_faces(vision_frame : VisionFrame, angle : Angle) -> Tuple[List[BoundingBox], List[Score], List[FaceLandmark5]]:
-	rotated_matrix, rotated_size = create_rotated_matrix_and_size(angle, vision_frame.shape[:2][::-1])
-	rotated_vision_frame = cv2.warpAffine(vision_frame, rotated_matrix, rotated_size)
-	rotated_inverse_matrix = cv2.invertAffineTransform(rotated_matrix)
-	bounding_boxes, face_scores, face_landmarks_5 = detect_faces(rotated_vision_frame)
-	bounding_boxes = [ transform_bounding_box(bounding_box, rotated_inverse_matrix) for bounding_box in bounding_boxes ]
-	face_landmarks_5 = [ transform_points(face_landmark_5, rotated_inverse_matrix) for face_landmark_5 in face_landmarks_5 ]
+def prepare_margin(vision_frame : VisionFrame) -> Margin:
+	margin_top = int(vision_frame.shape[0] * numpy.interp(state_manager.get_item('face_detector_margin')[0], [ 0, 100 ], [ 0, 0.5 ]))
+	margin_right = int(vision_frame.shape[1] * numpy.interp(state_manager.get_item('face_detector_margin')[1], [ 0, 100 ], [ 0, 0.5 ]))
+	margin_bottom = int(vision_frame.shape[0] * numpy.interp(state_manager.get_item('face_detector_margin')[2], [ 0, 100 ], [ 0, 0.5 ]))
+	margin_left = int(vision_frame.shape[1] * numpy.interp(state_manager.get_item('face_detector_margin')[3], [ 0, 100 ], [ 0, 0.5 ]))
+	return margin_top, margin_right, margin_bottom, margin_left
+
+
+def detect_faces_by_angle(vision_frame : VisionFrame, face_angle : Angle) -> Tuple[List[BoundingBox], List[Score], List[FaceLandmark5]]:
+	rotation_matrix, rotation_size = create_rotation_matrix_and_size(face_angle, vision_frame.shape[:2][::-1])
+	rotation_vision_frame = cv2.warpAffine(vision_frame, rotation_matrix, rotation_size)
+	rotation_inverse_matrix = cv2.invertAffineTransform(rotation_matrix)
+	bounding_boxes, face_scores, face_landmarks_5 = detect_faces(rotation_vision_frame)
+	bounding_boxes = [ transform_bounding_box(bounding_box, rotation_inverse_matrix) for bounding_box in bounding_boxes ]
+	face_landmarks_5 = [ transform_points(face_landmark_5, rotation_inverse_matrix) for face_landmark_5 in face_landmarks_5 ]
 	return bounding_boxes, face_scores, face_landmarks_5
 
 
@@ -162,7 +222,8 @@ def detect_with_retinaface(vision_frame : VisionFrame, face_detector_size : str)
 	detection = forward_with_retinaface(detect_vision_frame)
 
 	for index, feature_stride in enumerate(feature_strides):
-		keep_indices = numpy.where(detection[index] >= face_detector_score)[0]
+		face_scores_raw = detection[index]
+		keep_indices = numpy.where(face_scores_raw >= face_detector_score)[0]
 
 		if numpy.any(keep_indices):
 			stride_height = face_detector_height // feature_stride
@@ -180,7 +241,7 @@ def detect_with_retinaface(vision_frame : VisionFrame, face_detector_size : str)
 					bounding_box_raw[3] * ratio_height
 				]))
 
-			for face_score_raw in detection[index][keep_indices]:
+			for face_score_raw in face_scores_raw[keep_indices]:
 				face_scores.append(face_score_raw[0])
 
 			for face_landmark_raw_5 in distance_to_face_landmark_5(anchors, face_landmarks_5_raw)[keep_indices]:
@@ -206,7 +267,8 @@ def detect_with_scrfd(vision_frame : VisionFrame, face_detector_size : str) -> T
 	detection = forward_with_scrfd(detect_vision_frame)
 
 	for index, feature_stride in enumerate(feature_strides):
-		keep_indices = numpy.where(detection[index] >= face_detector_score)[0]
+		face_scores_raw = detection[index]
+		keep_indices = numpy.where(face_scores_raw >= face_detector_score)[0]
 
 		if numpy.any(keep_indices):
 			stride_height = face_detector_height // feature_stride
@@ -224,7 +286,7 @@ def detect_with_scrfd(vision_frame : VisionFrame, face_detector_size : str) -> T
 					bounding_box_raw[3] * ratio_height
 				]))
 
-			for face_score_raw in detection[index][keep_indices]:
+			for face_score_raw in face_scores_raw[keep_indices]:
 				face_scores.append(face_score_raw[0])
 
 			for face_landmark_raw_5 in distance_to_face_landmark_5(anchors, face_landmarks_5_raw)[keep_indices]:
@@ -271,6 +333,67 @@ def detect_with_yolo_face(vision_frame : VisionFrame, face_detector_size : str)
 	return bounding_boxes, face_scores, face_landmarks_5
 
 
+def detect_with_yunet(vision_frame : VisionFrame, face_detector_size : str) -> Tuple[List[BoundingBox], List[Score], List[FaceLandmark5]]:
+	bounding_boxes = []
+	face_scores = []
+	face_landmarks_5 = []
+	feature_strides = [ 8, 16, 32 ]
+	feature_map_channel = 3
+	anchor_total = 1
+	face_detector_score = state_manager.get_item('face_detector_score')
+	face_detector_width, face_detector_height = unpack_resolution(face_detector_size)
+	temp_vision_frame = restrict_frame(vision_frame, (face_detector_width, face_detector_height))
+	ratio_height = vision_frame.shape[0] / temp_vision_frame.shape[0]
+	ratio_width = vision_frame.shape[1] / temp_vision_frame.shape[1]
+	detect_vision_frame = prepare_detect_frame(temp_vision_frame, face_detector_size)
+	detect_vision_frame = normalize_detect_frame(detect_vision_frame, [ 0, 255 ])
+	detection = forward_with_yunet(detect_vision_frame)
+
+	for index, feature_stride in enumerate(feature_strides):
+		face_scores_raw = (detection[index] * detection[index + feature_map_channel]).reshape(-1)
+		keep_indices = numpy.where(face_scores_raw >= face_detector_score)[0]
+
+		if numpy.any(keep_indices):
+			stride_height = face_detector_height // feature_stride
+			stride_width = face_detector_width // feature_stride
+			anchors = create_static_anchors(feature_stride, anchor_total, stride_height, stride_width)
+			bounding_boxes_center = detection[index + feature_map_channel * 2].squeeze(0)[:, :2] * feature_stride + anchors
+			bounding_boxes_size = numpy.exp(detection[index + feature_map_channel * 2].squeeze(0)[:, 2:4]) * feature_stride
+			face_landmarks_5_raw = detection[index + feature_map_channel * 3].squeeze(0)
+
+			bounding_boxes_raw = numpy.stack(
+			[
+				bounding_boxes_center[:, 0] - bounding_boxes_size[:, 0] / 2,
+				bounding_boxes_center[:, 1] - bounding_boxes_size[:, 1] / 2,
+				bounding_boxes_center[:, 0] + bounding_boxes_size[:, 0] / 2,
+				bounding_boxes_center[:, 1] + bounding_boxes_size[:, 1] / 2
+			], axis = -1)
+
+			for bounding_box_raw in bounding_boxes_raw[keep_indices]:
+				bounding_boxes.append(numpy.array(
+				[
+					bounding_box_raw[0] * ratio_width,
+					bounding_box_raw[1] * ratio_height,
+					bounding_box_raw[2] * ratio_width,
+					bounding_box_raw[3] * ratio_height
+				]))
+
+			face_scores.extend(face_scores_raw[keep_indices])
+			face_landmarks_5_raw = numpy.concatenate(
+			[
+				face_landmarks_5_raw[:, [ 0, 1 ]] * feature_stride + anchors,
+				face_landmarks_5_raw[:, [ 2, 3 ]] * feature_stride + anchors,
+				face_landmarks_5_raw[:, [ 4, 5 ]] * feature_stride + anchors,
+				face_landmarks_5_raw[:, [ 6, 7 ]] * feature_stride + anchors,
+				face_landmarks_5_raw[:, [ 8, 9 ]] * feature_stride + anchors
+			], axis = -1).reshape(-1, 5, 2)
+
+			for face_landmark_raw_5 in face_landmarks_5_raw[keep_indices]:
+				face_landmarks_5.append(face_landmark_raw_5 * [ ratio_width, ratio_height ])
+
+	return bounding_boxes, face_scores, face_landmarks_5
+
+
 def forward_with_retinaface(detect_vision_frame : VisionFrame) -> Detection:
 	face_detector = get_inference_pool().get('retinaface')
 
@@ -307,6 +430,18 @@ def forward_with_yolo_face(detect_vision_frame : VisionFrame) -> Detection:
 	return detection
 
 
+def forward_with_yunet(detect_vision_frame : VisionFrame) -> Detection:
+	face_detector = get_inference_pool().get('yunet')
+
+	with thread_semaphore():
+		detection = face_detector.run(None,
+		{
+			'input': detect_vision_frame
+		})
+
+	return detection
+
+
 def prepare_detect_frame(temp_vision_frame : VisionFrame, face_detector_size : str) -> VisionFrame:
 	face_detector_width, face_detector_height = unpack_resolution(face_detector_size)
 	detect_vision_frame = numpy.zeros((face_detector_height, face_detector_width, 3))

facefusion/face_helper.py

@@ -69,8 +69,8 @@ WARP_TEMPLATE_SET : WarpTemplateSet =\
 
 
 def estimate_matrix_by_face_landmark_5(face_landmark_5 : FaceLandmark5, warp_template : WarpTemplate, crop_size : Size) -> Matrix:
-	normed_warp_template = WARP_TEMPLATE_SET.get(warp_template) * crop_size
-	affine_matrix = cv2.estimateAffinePartial2D(face_landmark_5, normed_warp_template, method = cv2.RANSAC, ransacReprojThreshold = 100)[0]
+	warp_template_norm = WARP_TEMPLATE_SET.get(warp_template) * crop_size
+	affine_matrix = cv2.estimateAffinePartial2D(face_landmark_5, warp_template_norm, method = cv2.RANSAC, ransacReprojThreshold = 100)[0]
 	return affine_matrix
 
 
@@ -81,7 +81,7 @@ def warp_face_by_face_landmark_5(temp_vision_frame : VisionFrame, face_landmark_
 
 
 def warp_face_by_bounding_box(temp_vision_frame : VisionFrame, bounding_box : BoundingBox, crop_size : Size) -> Tuple[VisionFrame, Matrix]:
-	source_points = numpy.array([ [ bounding_box[0], bounding_box[1] ], [bounding_box[2], bounding_box[1] ], [ bounding_box[0], bounding_box[3] ] ]).astype(numpy.float32)
+	source_points = numpy.array([ [ bounding_box[0], bounding_box[1] ], [ bounding_box[2], bounding_box[1] ], [ bounding_box[0], bounding_box[3] ] ]).astype(numpy.float32)
 	target_points = numpy.array([ [ 0, 0 ], [ crop_size[0], 0 ], [ 0, crop_size[1] ] ]).astype(numpy.float32)
 	affine_matrix = cv2.getAffineTransform(source_points, target_points)
 	if bounding_box[2] - bounding_box[0] > crop_size[0] or bounding_box[3] - bounding_box[1] > crop_size[1]:
@@ -98,39 +98,59 @@ def warp_face_by_translation(temp_vision_frame : VisionFrame, translation : Tran
 	return crop_vision_frame, affine_matrix
 
 
-def paste_back(temp_vision_frame : VisionFrame, crop_vision_frame : VisionFrame, crop_mask : Mask, affine_matrix : Matrix) -> VisionFrame:
+def paste_back(temp_vision_frame : VisionFrame, crop_vision_frame : VisionFrame, crop_vision_mask : Mask, affine_matrix : Matrix) -> VisionFrame:
+	paste_bounding_box, paste_matrix = calculate_paste_area(temp_vision_frame, crop_vision_frame, affine_matrix)
+	x1, y1, x2, y2 = paste_bounding_box
+	paste_width = x2 - x1
+	paste_height = y2 - y1
+	inverse_vision_mask = cv2.warpAffine(crop_vision_mask, paste_matrix, (paste_width, paste_height)).clip(0, 1)
+	inverse_vision_mask = numpy.expand_dims(inverse_vision_mask, axis = -1)
+	inverse_vision_frame = cv2.warpAffine(crop_vision_frame, paste_matrix, (paste_width, paste_height), borderMode = cv2.BORDER_REPLICATE)
+	temp_vision_frame = temp_vision_frame.copy()
+	paste_vision_frame = temp_vision_frame[y1:y2, x1:x2]
+	paste_vision_frame = paste_vision_frame * (1 - inverse_vision_mask) + inverse_vision_frame * inverse_vision_mask
+	temp_vision_frame[y1:y2, x1:x2] = paste_vision_frame.astype(temp_vision_frame.dtype)
+	return temp_vision_frame
+
+
+def calculate_paste_area(temp_vision_frame : VisionFrame, crop_vision_frame : VisionFrame, affine_matrix : Matrix) -> Tuple[BoundingBox, Matrix]:
+	temp_height, temp_width = temp_vision_frame.shape[:2]
+	crop_height, crop_width = crop_vision_frame.shape[:2]
 	inverse_matrix = cv2.invertAffineTransform(affine_matrix)
-	temp_size = temp_vision_frame.shape[:2][::-1]
-	inverse_mask = cv2.warpAffine(crop_mask, inverse_matrix, temp_size).clip(0, 1)
-	inverse_vision_frame = cv2.warpAffine(crop_vision_frame, inverse_matrix, temp_size, borderMode = cv2.BORDER_REPLICATE)
-	paste_vision_frame = temp_vision_frame.copy()
-	paste_vision_frame[:, :, 0] = inverse_mask * inverse_vision_frame[:, :, 0] + (1 - inverse_mask) * temp_vision_frame[:, :, 0]
-	paste_vision_frame[:, :, 1] = inverse_mask * inverse_vision_frame[:, :, 1] + (1 - inverse_mask) * temp_vision_frame[:, :, 1]
-	paste_vision_frame[:, :, 2] = inverse_mask * inverse_vision_frame[:, :, 2] + (1 - inverse_mask) * temp_vision_frame[:, :, 2]
-	return paste_vision_frame
+	crop_points = numpy.array([ [ 0, 0 ], [ crop_width, 0 ], [ crop_width, crop_height ], [ 0, crop_height ] ])
+	paste_region_points = transform_points(crop_points, inverse_matrix)
+	paste_region_point_min = numpy.floor(paste_region_points.min(axis = 0)).astype(int)
+	paste_region_point_max = numpy.ceil(paste_region_points.max(axis = 0)).astype(int)
+	x1, y1 = numpy.clip(paste_region_point_min, 0, [ temp_width, temp_height ])
+	x2, y2 = numpy.clip(paste_region_point_max, 0, [ temp_width, temp_height ])
+	paste_bounding_box = numpy.array([ x1, y1, x2, y2 ])
+	paste_matrix = inverse_matrix.copy()
+	paste_matrix[0, 2] -= x1
+	paste_matrix[1, 2] -= y1
+	return paste_bounding_box, paste_matrix
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_anchors(feature_stride : int, anchor_total : int, stride_height : int, stride_width : int) -> Anchors:
-	y, x = numpy.mgrid[:stride_height, :stride_width][::-1]
+	x, y = numpy.mgrid[:stride_width, :stride_height]
 	anchors = numpy.stack((y, x), axis = -1)
 	anchors = (anchors * feature_stride).reshape((-1, 2))
 	anchors = numpy.stack([ anchors ] * anchor_total, axis = 1).reshape((-1, 2))
 	return anchors
 
 
-def create_rotated_matrix_and_size(angle : Angle, size : Size) -> Tuple[Matrix, Size]:
-	rotated_matrix = cv2.getRotationMatrix2D((size[0] / 2, size[1] / 2), angle, 1)
-	rotated_size = numpy.dot(numpy.abs(rotated_matrix[:, :2]), size)
-	rotated_matrix[:, -1] += (rotated_size - size) * 0.5 #type:ignore[misc]
-	rotated_size = int(rotated_size[0]), int(rotated_size[1])
-	return rotated_matrix, rotated_size
+def create_rotation_matrix_and_size(angle : Angle, size : Size) -> Tuple[Matrix, Size]:
+	rotation_matrix = cv2.getRotationMatrix2D((size[0] / 2, size[1] / 2), angle, 1)
+	rotation_size = numpy.dot(numpy.abs(rotation_matrix[:, :2]), size)
+	rotation_matrix[:, -1] += (rotation_size - size) * 0.5 #type:ignore[misc]
+	rotation_size = int(rotation_size[0]), int(rotation_size[1])
+	return rotation_matrix, rotation_size
 
 
 def create_bounding_box(face_landmark_68 : FaceLandmark68) -> BoundingBox:
-	min_x, min_y = numpy.min(face_landmark_68, axis = 0)
-	max_x, max_y = numpy.max(face_landmark_68, axis = 0)
-	bounding_box = normalize_bounding_box(numpy.array([ min_x, min_y, max_x, max_y ]))
+	x1, y1 = numpy.min(face_landmark_68, axis = 0)
+	x2, y2 = numpy.max(face_landmark_68, axis = 0)
+	bounding_box = normalize_bounding_box(numpy.array([ x1, y1, x2, y2 ]))
 	return bounding_box
 
 
@@ -209,8 +229,8 @@ def estimate_face_angle(face_landmark_68 : FaceLandmark68) -> Angle:
 
 
 def apply_nms(bounding_boxes : List[BoundingBox], scores : List[Score], score_threshold : float, nms_threshold : float) -> Sequence[int]:
-	normed_bounding_boxes = [ (x1, y1, x2 - x1, y2 - y1) for (x1, y1, x2, y2) in bounding_boxes ]
-	keep_indices = cv2.dnn.NMSBoxes(normed_bounding_boxes, scores, score_threshold = score_threshold, nms_threshold = nms_threshold)
+	bounding_boxes_norm = [ (x1, y1, x2 - x1, y2 - y1) for (x1, y1, x2, y2) in bounding_boxes ]
+	keep_indices = cv2.dnn.NMSBoxes(bounding_boxes_norm, scores, score_threshold = score_threshold, nms_threshold = nms_threshold)
 	return keep_indices
 
 
@@ -226,9 +246,31 @@ def get_nms_threshold(face_detector_model : FaceDetectorModel, face_detector_ang
 	return 0.4
 
 
-def merge_matrix(matrices : List[Matrix]) -> Matrix:
-	merged_matrix = numpy.vstack([ matrices[0], [ 0, 0, 1 ] ])
-	for matrix in matrices[1:]:
-		matrix = numpy.vstack([ matrix, [ 0, 0, 1 ] ])
-		merged_matrix = numpy.dot(merged_matrix, matrix)
-	return merged_matrix[:2, :]
+def merge_matrix(temp_matrices : List[Matrix]) -> Matrix:
+	matrix = numpy.vstack([ temp_matrices[0], [ 0, 0, 1 ] ])
+
+	for temp_matrix in temp_matrices[1:]:
+		temp_matrix = numpy.vstack([ temp_matrix, [ 0, 0, 1 ] ])
+		matrix = numpy.dot(temp_matrix, matrix)
+
+	return matrix[:2, :]
+
+
+def calculate_bounding_box_overlap(bounding_box_a : BoundingBox, bounding_box_b : BoundingBox) -> float:
+	intersection_x1 = max(bounding_box_a[0], bounding_box_b[0])
+	intersection_y1 = max(bounding_box_a[1], bounding_box_b[1])
+	intersection_x2 = min(bounding_box_a[2], bounding_box_b[2])
+	intersection_y2 = min(bounding_box_a[3], bounding_box_b[3])
+	intersection = max(0, intersection_x2 - intersection_x1) * max(0, intersection_y2 - intersection_y1)
+	bounding_box_area = (bounding_box_a[2] - bounding_box_a[0]) * (bounding_box_a[3] - bounding_box_a[1])
+	reference_bounding_box_area = (bounding_box_b[2] - bounding_box_b[0]) * (bounding_box_b[3] - bounding_box_b[1])
+	union = bounding_box_area + reference_bounding_box_area - intersection
+
+	if union > 0:
+		return intersection / union
+
+	return 0.0
+
+
+def average_points(points_previous : Points, points_next : Points, average_factor : float) -> Points:
+	return points_previous * (1 - average_factor) + points_next * average_factor

facefusion/face_landmarker.py

@@ -6,18 +6,24 @@ import numpy
 
 from facefusion import inference_manager, state_manager
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.face_helper import create_rotated_matrix_and_size, estimate_matrix_by_face_landmark_5, transform_points, warp_face_by_translation
+from facefusion.face_helper import create_rotation_matrix_and_size, estimate_matrix_by_face_landmark_5, transform_points, warp_face_by_translation
 from facefusion.filesystem import resolve_relative_path
 from facefusion.thread_helper import conditional_thread_semaphore
 from facefusion.types import Angle, BoundingBox, DownloadScope, DownloadSet, FaceLandmark5, FaceLandmark68, InferencePool, ModelSet, Prediction, Score, VisionFrame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'2dfan4':
 		{
+			'__metadata__':
+			{
+				'vendor': 'breadbread1984',
+				'license': 'MIT',
+				'year': 2018
+			},
 			'hashes':
 			{
 				'2dfan4':
@@ -38,6 +44,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'peppa_wutz':
 		{
+			'__metadata__':
+			{
+				'vendor': 'Unknown',
+				'license': 'Apache-2.0',
+				'year': 2023
+			},
 			'hashes':
 			{
 				'peppa_wutz':
@@ -58,6 +70,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'fan_68_5':
 		{
+			'__metadata__':
+			{
+				'vendor': 'FaceFusion',
+				'license': 'OpenRAIL-M',
+				'year': 2024
+			},
 			'hashes':
 			{
 				'fan_68_5':
@@ -136,14 +154,14 @@ def detect_with_2dfan4(temp_vision_frame: VisionFrame, bounding_box: BoundingBox
 	model_size = create_static_model_set('full').get('2dfan4').get('size')
 	scale = 195 / numpy.subtract(bounding_box[2:], bounding_box[:2]).max().clip(1, None)
 	translation = (model_size[0] - numpy.add(bounding_box[2:], bounding_box[:2]) * scale) * 0.5
-	rotated_matrix, rotated_size = create_rotated_matrix_and_size(face_angle, model_size)
+	rotation_matrix, rotation_size = create_rotation_matrix_and_size(face_angle, model_size)
 	crop_vision_frame, affine_matrix = warp_face_by_translation(temp_vision_frame, translation, scale, model_size)
-	crop_vision_frame = cv2.warpAffine(crop_vision_frame, rotated_matrix, rotated_size)
+	crop_vision_frame = cv2.warpAffine(crop_vision_frame, rotation_matrix, rotation_size)
 	crop_vision_frame = conditional_optimize_contrast(crop_vision_frame)
 	crop_vision_frame = crop_vision_frame.transpose(2, 0, 1).astype(numpy.float32) / 255.0
 	face_landmark_68, face_heatmap = forward_with_2dfan4(crop_vision_frame)
 	face_landmark_68 = face_landmark_68[:, :, :2][0] / 64 * 256
-	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(rotated_matrix))
+	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(rotation_matrix))
 	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(affine_matrix))
 	face_landmark_score_68 = numpy.amax(face_heatmap, axis = (2, 3))
 	face_landmark_score_68 = numpy.mean(face_landmark_score_68)
@@ -155,15 +173,15 @@ def detect_with_peppa_wutz(temp_vision_frame : VisionFrame, bounding_box : Bound
 	model_size = create_static_model_set('full').get('peppa_wutz').get('size')
 	scale = 195 / numpy.subtract(bounding_box[2:], bounding_box[:2]).max().clip(1, None)
 	translation = (model_size[0] - numpy.add(bounding_box[2:], bounding_box[:2]) * scale) * 0.5
-	rotated_matrix, rotated_size = create_rotated_matrix_and_size(face_angle, model_size)
+	rotation_matrix, rotation_size = create_rotation_matrix_and_size(face_angle, model_size)
 	crop_vision_frame, affine_matrix = warp_face_by_translation(temp_vision_frame, translation, scale, model_size)
-	crop_vision_frame = cv2.warpAffine(crop_vision_frame, rotated_matrix, rotated_size)
+	crop_vision_frame = cv2.warpAffine(crop_vision_frame, rotation_matrix, rotation_size)
 	crop_vision_frame = conditional_optimize_contrast(crop_vision_frame)
 	crop_vision_frame = crop_vision_frame.transpose(2, 0, 1).astype(numpy.float32) / 255.0
 	crop_vision_frame = numpy.expand_dims(crop_vision_frame, axis = 0)
 	prediction = forward_with_peppa_wutz(crop_vision_frame)
 	face_landmark_68 = prediction.reshape(-1, 3)[:, :2] / 64 * model_size[0]
-	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(rotated_matrix))
+	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(rotation_matrix))
 	face_landmark_68 = transform_points(face_landmark_68, cv2.invertAffineTransform(affine_matrix))
 	face_landmark_score_68 = prediction.reshape(-1, 3)[:, 2].mean()
 	face_landmark_score_68 = numpy.interp(face_landmark_score_68, [ 0, 0.95 ], [ 0, 1 ])

facefusion/face_masker.py

@@ -3,22 +3,27 @@ from typing import List, Tuple
 
 import cv2
 import numpy
-from cv2.typing import Size
 
 import facefusion.choices
 from facefusion import inference_manager, state_manager
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
 from facefusion.filesystem import resolve_relative_path
 from facefusion.thread_helper import conditional_thread_semaphore
-from facefusion.types import DownloadScope, DownloadSet, FaceLandmark68, FaceMaskRegion, InferencePool, Mask, ModelSet, Padding, VisionFrame
+from facefusion.types import DownloadScope, DownloadSet, FaceLandmark68, FaceMaskArea, FaceMaskRegion, InferencePool, Mask, ModelSet, Padding, VisionFrame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'xseg_1':
 		{
+			'__metadata__':
+			{
+				'vendor': 'DeepFaceLab',
+				'license': 'GPL-3.0',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_occluder':
@@ -39,6 +44,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'xseg_2':
 		{
+			'__metadata__':
+			{
+				'vendor': 'DeepFaceLab',
+				'license': 'GPL-3.0',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_occluder':
@@ -59,6 +70,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'xseg_3':
 		{
+			'__metadata__':
+			{
+				'vendor': 'DeepFaceLab',
+				'license': 'GPL-3.0',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_occluder':
@@ -79,6 +96,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'bisenet_resnet_18':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yakhyo',
+				'license': 'MIT',
+				'year': 2024
+			},
 			'hashes':
 			{
 				'face_parser':
@@ -99,6 +122,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'bisenet_resnet_34':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yakhyo',
+				'license': 'MIT',
+				'year': 2024
+			},
 			'hashes':
 			{
 				'face_parser':
@@ -121,7 +150,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 
 
 def get_inference_pool() -> InferencePool:
-	model_names = [state_manager.get_item('face_occluder_model'), state_manager.get_item('face_parser_model')]
+	model_names = [ state_manager.get_item('face_occluder_model'), state_manager.get_item('face_parser_model') ]
 	_, model_source_set = collect_model_downloads()
 
 	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
@@ -138,7 +167,7 @@ def collect_model_downloads() -> Tuple[DownloadSet, DownloadSet]:
 	model_source_set = {}
 
 	for face_occluder_model in [ 'xseg_1', 'xseg_2', 'xseg_3' ]:
-		if state_manager.get_item('face_occluder_model') == face_occluder_model:
+		if state_manager.get_item('face_occluder_model') in [ 'many', face_occluder_model ]:
 			model_hash_set[face_occluder_model] = model_set.get(face_occluder_model).get('hashes').get('face_occluder')
 			model_source_set[face_occluder_model] = model_set.get(face_occluder_model).get('sources').get('face_occluder')
 
@@ -156,8 +185,8 @@ def pre_check() -> bool:
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
 
-@lru_cache(maxsize = None)
-def create_static_box_mask(crop_size : Size, face_mask_blur : float, face_mask_padding : Padding) -> Mask:
+def create_box_mask(crop_vision_frame : VisionFrame, face_mask_blur : float, face_mask_padding : Padding) -> Mask:
+	crop_size = crop_vision_frame.shape[:2][::-1]
 	blur_amount = int(crop_size[0] * 0.5 * face_mask_blur)
 	blur_area = max(blur_amount // 2, 1)
 	box_mask : Mask = numpy.ones(crop_size).astype(numpy.float32)
@@ -165,24 +194,50 @@ def create_static_box_mask(crop_size : Size, face_mask_blur : float, face_mask_p
 	box_mask[-max(blur_area, int(crop_size[1] * face_mask_padding[2] / 100)):, :] = 0
 	box_mask[:, :max(blur_area, int(crop_size[0] * face_mask_padding[3] / 100))] = 0
 	box_mask[:, -max(blur_area, int(crop_size[0] * face_mask_padding[1] / 100)):] = 0
+
 	if blur_amount > 0:
 		box_mask = cv2.GaussianBlur(box_mask, (0, 0), blur_amount * 0.25)
 	return box_mask
 
 
 def create_occlusion_mask(crop_vision_frame : VisionFrame) -> Mask:
-	model_name = state_manager.get_item('face_occluder_model')
-	model_size = create_static_model_set('full').get(model_name).get('size')
-	prepare_vision_frame = cv2.resize(crop_vision_frame, model_size)
-	prepare_vision_frame = numpy.expand_dims(prepare_vision_frame, axis = 0).astype(numpy.float32) / 255.0
-	prepare_vision_frame = prepare_vision_frame.transpose(0, 1, 2, 3)
-	occlusion_mask = forward_occlude_face(prepare_vision_frame)
-	occlusion_mask = occlusion_mask.transpose(0, 1, 2).clip(0, 1).astype(numpy.float32)
-	occlusion_mask = cv2.resize(occlusion_mask, crop_vision_frame.shape[:2][::-1])
+	temp_masks = []
+
+	if state_manager.get_item('face_occluder_model') == 'many':
+		model_names = [ 'xseg_1', 'xseg_2', 'xseg_3' ]
+	else:
+		model_names = [ state_manager.get_item('face_occluder_model') ]
+
+	for model_name in model_names:
+		model_size = create_static_model_set('full').get(model_name).get('size')
+		prepare_vision_frame = cv2.resize(crop_vision_frame, model_size)
+		prepare_vision_frame = numpy.expand_dims(prepare_vision_frame, axis = 0).astype(numpy.float32) / 255.0
+		prepare_vision_frame = prepare_vision_frame.transpose(0, 1, 2, 3)
+		temp_mask = forward_occlude_face(prepare_vision_frame, model_name)
+		temp_mask = temp_mask.transpose(0, 1, 2).clip(0, 1).astype(numpy.float32)
+		temp_mask = cv2.resize(temp_mask, crop_vision_frame.shape[:2][::-1])
+		temp_masks.append(temp_mask)
+
+	occlusion_mask = numpy.minimum.reduce(temp_masks)
 	occlusion_mask = (cv2.GaussianBlur(occlusion_mask.clip(0, 1), (0, 0), 5).clip(0.5, 1) - 0.5) * 2
 	return occlusion_mask
 
 
+def create_area_mask(crop_vision_frame : VisionFrame, face_landmark_68 : FaceLandmark68, face_mask_areas : List[FaceMaskArea]) -> Mask:
+	crop_size = crop_vision_frame.shape[:2][::-1]
+	landmark_points = []
+
+	for face_mask_area in face_mask_areas:
+		if face_mask_area in facefusion.choices.face_mask_area_set:
+			landmark_points.extend(facefusion.choices.face_mask_area_set.get(face_mask_area))
+
+	convex_hull = cv2.convexHull(face_landmark_68[landmark_points].astype(numpy.int32))
+	area_mask = numpy.zeros(crop_size).astype(numpy.float32)
+	cv2.fillConvexPoly(area_mask, convex_hull, 1.0) #type:ignore[call-overload]
+	area_mask = (cv2.GaussianBlur(area_mask.clip(0, 1), (0, 0), 5).clip(0.5, 1) - 0.5) * 2
+	return area_mask
+
+
 def create_region_mask(crop_vision_frame : VisionFrame, face_mask_regions : List[FaceMaskRegion]) -> Mask:
 	model_name = state_manager.get_item('face_parser_model')
 	model_size = create_static_model_set('full').get(model_name).get('size')
@@ -199,17 +254,7 @@ def create_region_mask(crop_vision_frame : VisionFrame, face_mask_regions : List
 	return region_mask
 
 
-def create_mouth_mask(face_landmark_68 : FaceLandmark68) -> Mask:
-	convex_hull = cv2.convexHull(face_landmark_68[numpy.r_[3:14, 31:36]].astype(numpy.int32))
-	mouth_mask : Mask = numpy.zeros((512, 512)).astype(numpy.float32)
-	mouth_mask = cv2.fillConvexPoly(mouth_mask, convex_hull, 1.0) #type:ignore[call-overload]
-	mouth_mask = cv2.erode(mouth_mask.clip(0, 1), numpy.ones((21, 3)))
-	mouth_mask = cv2.GaussianBlur(mouth_mask, (0, 0), sigmaX = 1, sigmaY = 15)
-	return mouth_mask
-
-
-def forward_occlude_face(prepare_vision_frame : VisionFrame) -> Mask:
-	model_name = state_manager.get_item('face_occluder_model')
+def forward_occlude_face(prepare_vision_frame : VisionFrame, model_name : str) -> Mask:
 	face_occluder = get_inference_pool().get(model_name)
 
 	with conditional_thread_semaphore():

facefusion/face_recognizer.py

@@ -11,12 +11,18 @@ from facefusion.thread_helper import conditional_thread_semaphore
 from facefusion.types import DownloadScope, Embedding, FaceLandmark5, InferencePool, ModelOptions, ModelSet, VisionFrame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'arcface':
 		{
+			'__metadata__':
+			{
+				'vendor': 'InsightFace',
+				'license': 'Non-Commercial',
+				'year': 2018
+			},
 			'hashes':
 			{
 				'face_recognizer':
@@ -62,26 +68,26 @@ def pre_check() -> bool:
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
 
-def calc_embedding(temp_vision_frame : VisionFrame, face_landmark_5 : FaceLandmark5) -> Tuple[Embedding, Embedding]:
+def calculate_face_embedding(temp_vision_frame : VisionFrame, face_landmark_5 : FaceLandmark5) -> Tuple[Embedding, Embedding]:
 	model_template = get_model_options().get('template')
 	model_size = get_model_options().get('size')
 	crop_vision_frame, matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5, model_template, model_size)
 	crop_vision_frame = crop_vision_frame / 127.5 - 1
 	crop_vision_frame = crop_vision_frame[:, :, ::-1].transpose(2, 0, 1).astype(numpy.float32)
 	crop_vision_frame = numpy.expand_dims(crop_vision_frame, axis = 0)
-	embedding = forward(crop_vision_frame)
-	embedding = embedding.ravel()
-	normed_embedding = embedding / numpy.linalg.norm(embedding)
-	return embedding, normed_embedding
+	face_embedding = forward(crop_vision_frame)
+	face_embedding = face_embedding.ravel()
+	face_embedding_norm = face_embedding / numpy.linalg.norm(face_embedding)
+	return face_embedding, face_embedding_norm
 
 
 def forward(crop_vision_frame : VisionFrame) -> Embedding:
 	face_recognizer = get_inference_pool().get('face_recognizer')
 
 	with conditional_thread_semaphore():
-		embedding = face_recognizer.run(None,
+		face_embedding = face_recognizer.run(None,
 		{
 			'input': crop_vision_frame
 		})[0]
 
-	return embedding
+	return face_embedding

facefusion/face_selector.py

@@ -2,68 +2,131 @@ from typing import List
 
 import numpy
 
+import facefusion.choices
 from facefusion import state_manager
-from facefusion.types import Face, FaceSelectorOrder, FaceSet, Gender, Race
+from facefusion.common_helper import get_first, get_middle
+from facefusion.face_creator import get_one_face, get_static_faces
+from facefusion.face_tracker import track_faces
+from facefusion.types import Face, FaceSelectorOrder, Gender, Race, Score, VisionFrame
 
 
-def find_similar_faces(faces : List[Face], reference_faces : FaceSet, face_distance : float) -> List[Face]:
-	similar_faces : List[Face] = []
+def select_faces(reference_vision_frame : VisionFrame, source_vision_frames : List[VisionFrame], target_vision_frames : List[VisionFrame]) -> List[Face]:
+	source_faces = get_static_faces(source_vision_frames)
 
-	if faces and reference_faces:
-		for reference_set in reference_faces:
-			if not similar_faces:
-				for reference_face in reference_faces[reference_set]:
-					for face in faces:
-						if compare_faces(face, reference_face, face_distance):
-							similar_faces.append(face)
-	return similar_faces
+	if state_manager.get_item('face_tracker_score') > 0:
+		target_faces = track_faces(target_vision_frames, state_manager.get_item('face_tracker_score'))
+	else:
+		target_faces = get_static_faces([ get_middle(target_vision_frames) ])
+
+	if state_manager.get_item('face_selector_mode') == 'many':
+		return sort_and_filter_faces(source_faces, target_faces)
+
+	if state_manager.get_item('face_selector_mode') == 'one':
+		target_face = get_one_face(sort_and_filter_faces(source_faces, target_faces))
+		if target_face:
+			return [ target_face ]
+
+	if state_manager.get_item('face_selector_mode') == 'reference':
+		reference_faces = get_static_faces([ reference_vision_frame ])
+		reference_faces = sort_and_filter_faces(source_faces, reference_faces)
+		reference_face = get_one_face(reference_faces, state_manager.get_item('reference_face_position'))
+
+		if reference_face:
+			match_faces = find_match_faces([ reference_face ], target_faces, state_manager.get_item('reference_face_distance'))
+			return match_faces
+
+	return []
+
+
+def find_match_faces(reference_faces : List[Face], target_faces : List[Face], face_distance : float) -> List[Face]:
+	match_faces : List[Face] = []
+
+	for reference_face in reference_faces:
+		if reference_face:
+			for index, target_face in enumerate(target_faces):
+				if compare_faces(target_face, reference_face, face_distance):
+					match_faces.append(target_faces[index])
+
+	return match_faces
 
 
 def compare_faces(face : Face, reference_face : Face, face_distance : float) -> bool:
-	current_face_distance = calc_face_distance(face, reference_face)
+	current_face_distance = calculate_face_distance(face, reference_face)
 	current_face_distance = float(numpy.interp(current_face_distance, [ 0, 2 ], [ 0, 1 ]))
 	return current_face_distance < face_distance
 
 
-def calc_face_distance(face : Face, reference_face : Face) -> float:
-	if hasattr(face, 'normed_embedding') and hasattr(reference_face, 'normed_embedding'):
-		return 1 - numpy.dot(face.normed_embedding, reference_face.normed_embedding)
+def calculate_face_distance(face : Face, reference_face : Face) -> float:
+	if hasattr(face, 'embedding_norm') and hasattr(reference_face, 'embedding_norm'):
+		return 1 - numpy.dot(face.embedding_norm, reference_face.embedding_norm)
 	return 0
 
 
-def sort_and_filter_faces(faces : List[Face]) -> List[Face]:
-	if faces:
+def sort_and_filter_faces(source_faces : List[Face], target_faces : List[Face]) -> List[Face]:
+	if target_faces:
 		if state_manager.get_item('face_selector_order'):
-			faces = sort_faces_by_order(faces, state_manager.get_item('face_selector_order'))
-		if state_manager.get_item('face_selector_gender'):
-			faces = filter_faces_by_gender(faces, state_manager.get_item('face_selector_gender'))
-		if state_manager.get_item('face_selector_race'):
-			faces = filter_faces_by_race(faces, state_manager.get_item('face_selector_race'))
+			target_faces = sort_faces_by_order(target_faces, state_manager.get_item('face_selector_order'))
+
+		face_selector_gender = state_manager.get_item('face_selector_gender')
+		face_selector_race = state_manager.get_item('face_selector_race')
+
+		if source_faces and face_selector_gender == 'auto' or face_selector_race == 'auto':
+			source_face = get_first(sort_faces_by_order(source_faces, 'large-small'))
+
+			if source_face:
+				if face_selector_gender == 'auto':
+					face_selector_gender = source_face.gender
+				if face_selector_race == 'auto':
+					face_selector_race = source_face.race
+
+		if face_selector_gender in facefusion.choices.genders:
+			target_faces = filter_faces_by_gender(target_faces, face_selector_gender)
+
+		if face_selector_race in facefusion.choices.races:
+			target_faces = filter_faces_by_race(target_faces, face_selector_race)
+
 		if state_manager.get_item('face_selector_age_start') or state_manager.get_item('face_selector_age_end'):
-			faces = filter_faces_by_age(faces, state_manager.get_item('face_selector_age_start'), state_manager.get_item('face_selector_age_end'))
-	return faces
+			target_faces = filter_faces_by_age(target_faces, state_manager.get_item('face_selector_age_start'), state_manager.get_item('face_selector_age_end'))
+
+	return target_faces
 
 
 def sort_faces_by_order(faces : List[Face], order : FaceSelectorOrder) -> List[Face]:
 	if order == 'left-right':
-		return sorted(faces, key = lambda face: face.bounding_box[0])
+		return sorted(faces, key = get_bounding_box_left)
 	if order == 'right-left':
-		return sorted(faces, key = lambda face: face.bounding_box[0], reverse = True)
+		return sorted(faces, key = get_bounding_box_left, reverse = True)
 	if order == 'top-bottom':
-		return sorted(faces, key = lambda face: face.bounding_box[1])
+		return sorted(faces, key = get_bounding_box_top)
 	if order == 'bottom-top':
-		return sorted(faces, key = lambda face: face.bounding_box[1], reverse = True)
+		return sorted(faces, key = get_bounding_box_top, reverse = True)
 	if order == 'small-large':
-		return sorted(faces, key = lambda face: (face.bounding_box[2] - face.bounding_box[0]) * (face.bounding_box[3] - face.bounding_box[1]))
+		return sorted(faces, key = get_bounding_box_area)
 	if order == 'large-small':
-		return sorted(faces, key = lambda face: (face.bounding_box[2] - face.bounding_box[0]) * (face.bounding_box[3] - face.bounding_box[1]), reverse = True)
+		return sorted(faces, key = get_bounding_box_area, reverse = True)
 	if order == 'best-worst':
-		return sorted(faces, key = lambda face: face.score_set.get('detector'), reverse = True)
+		return sorted(faces, key = get_face_detector_score, reverse = True)
 	if order == 'worst-best':
-		return sorted(faces, key = lambda face: face.score_set.get('detector'))
+		return sorted(faces, key = get_face_detector_score)
 	return faces
 
 
+def get_bounding_box_left(face : Face) -> float:
+	return face.bounding_box[0]
+
+
+def get_bounding_box_top(face : Face) -> float:
+	return face.bounding_box[1]
+
+
+def get_bounding_box_area(face : Face) -> float:
+	return (face.bounding_box[2] - face.bounding_box[0]) * (face.bounding_box[3] - face.bounding_box[1])
+
+
+def get_face_detector_score(face : Face) -> Score:
+	return face.score_set.get('detector')
+
+
 def filter_faces_by_gender(faces : List[Face], gender : Gender) -> List[Face]:
 	filter_faces = []
 

facefusion/face_store.py

@@ -1,56 +1,42 @@
-import hashlib
+import threading
 from typing import List, Optional
 
 import numpy
 
-from facefusion.types import Face, FaceSet, FaceStore, VisionFrame
+from facefusion.hash_helper import create_hash
+from facefusion.types import Face, FaceStore, VisionFrame
 
-FACE_STORE : FaceStore =\
-{
-	'static_faces': {},
-	'reference_faces': {}
-}
+FACE_STORE : FaceStore = {}
 
 
-def get_face_store() -> FaceStore:
-	return FACE_STORE
-
-
-def get_static_faces(vision_frame : VisionFrame) -> Optional[List[Face]]:
-	frame_hash = create_frame_hash(vision_frame)
-	if frame_hash in FACE_STORE['static_faces']:
-		return FACE_STORE['static_faces'][frame_hash]
-	return None
-
-
-def set_static_faces(vision_frame : VisionFrame, faces : List[Face]) -> None:
-	frame_hash = create_frame_hash(vision_frame)
-	if frame_hash:
-		FACE_STORE['static_faces'][frame_hash] = faces
-
-
-def clear_static_faces() -> None:
-	FACE_STORE['static_faces'] = {}
-
-
-def create_frame_hash(vision_frame : VisionFrame) -> Optional[str]:
+def get_faces(vision_frame : VisionFrame) -> Optional[List[Face]]:
 	if numpy.any(vision_frame):
-		frame_hash = hashlib.blake2b(vision_frame.tobytes(), digest_size = 16).hexdigest()
-		return frame_hash
+		vision_hash = create_hash(vision_frame.tobytes())
+
+		if FACE_STORE.get(vision_hash):
+			return FACE_STORE.get(vision_hash).get('faces')
+
 	return None
 
 
-def get_reference_faces() -> Optional[FaceSet]:
-	if FACE_STORE['reference_faces']:
-		return FACE_STORE['reference_faces']
-	return None
+def set_faces(vision_frame : VisionFrame, faces : List[Face]) -> None:
+	if numpy.any(vision_frame):
+		vision_hash = create_hash(vision_frame.tobytes())
+		FACE_STORE.setdefault(vision_hash,
+		{
+			'lock': threading.Lock()
+		})['faces'] = faces
 
 
-def append_reference_face(name : str, face : Face) -> None:
-	if name not in FACE_STORE['reference_faces']:
-		FACE_STORE['reference_faces'][name] = []
-	FACE_STORE['reference_faces'][name].append(face)
+def resolve_lock(vision_frame : VisionFrame) -> threading.Lock:
+	if numpy.any(vision_frame):
+		vision_hash = create_hash(vision_frame.tobytes())
+		return FACE_STORE.setdefault(vision_hash,
+		{
+			'lock': threading.Lock()
+		}).get('lock')
+	return threading.Lock()
 
 
-def clear_reference_faces() -> None:
-	FACE_STORE['reference_faces'] = {}
+def clear_faces() -> None:
+	FACE_STORE.clear()

facefusion/face_tracker.py

@@ -0,0 +1,61 @@
+from typing import List
+
+from facefusion.common_helper import get_first, get_last
+from facefusion.face_creator import get_static_faces, refill_faces
+from facefusion.face_helper import calculate_bounding_box_overlap
+from facefusion.types import Face, FaceTrack, Score, VisionFrame
+
+
+def track_faces(vision_frames : List[VisionFrame], score : Score) -> List[Face]:
+	target_index = len(vision_frames) // 2
+	face_tracks = create_face_tracks(vision_frames, score)
+	temp_faces = []
+
+	for face_track in face_tracks:
+		track_indices = sorted(face_track)
+		track_index_first = get_first(track_indices)
+		track_index_last = get_last(track_indices)
+		track_range = range(track_index_first, track_index_last + 1)
+
+		if target_index in track_range:
+			fill_faces = []
+
+			for index in track_range:
+				fill_faces.append(face_track.get(index))
+
+			temp_faces.append(refill_faces(fill_faces)[target_index - track_index_first])
+
+	return temp_faces
+
+
+def create_face_tracks(vision_frames : List[VisionFrame], score : Score) -> List[FaceTrack]:
+	face_tracks : List[FaceTrack] = []
+
+	for frame_index, vision_frame in enumerate(vision_frames):
+		for face in get_static_faces([ vision_frame ]):
+			face_track = select_face_track(face_tracks, face, score)
+
+			if face_track:
+				face_track[frame_index] = face
+			else:
+				face_tracks.append(
+				{
+					frame_index : face
+				})
+
+	return face_tracks
+
+
+def select_face_track(face_tracks : List[FaceTrack], face : Face, score : Score) -> FaceTrack:
+	select_track : FaceTrack = {}
+	select_score = score
+
+	for face_track in face_tracks:
+		track_face = face_track.get(get_last(face_track))
+		track_score = calculate_bounding_box_overlap(face.bounding_box, track_face.bounding_box)
+
+		if track_score > select_score:
+			select_score = track_score
+			select_track = face_track
+
+	return select_track

facefusion/ffmpeg.py

@@ -1,19 +1,19 @@
 import os
 import subprocess
 import tempfile
-from typing import List, Optional
+from functools import partial
+from typing import List, Optional, cast
 
 from tqdm import tqdm
 
 import facefusion.choices
-from facefusion import ffmpeg_builder, logger, process_manager, state_manager, wording
+from facefusion import ffmpeg_builder, ffprobe, logger, process_manager, state_manager, translator, vision
 from facefusion.filesystem import get_file_format, remove_file
-from facefusion.temp_helper import get_temp_file_path, get_temp_frames_pattern
-from facefusion.types import AudioBuffer, Commands, EncoderSet, Fps, UpdateProgress
-from facefusion.vision import detect_video_duration, detect_video_fps, predict_video_frame_total
+from facefusion.temp_helper import get_temp_file_path, get_temp_frame_pattern
+from facefusion.types import AudioBuffer, AudioEncoder, Command, EncoderSet, Fps, Resolution, UpdateProgress, VideoEncoder, VideoFormat, VideoReaderMetadata
 
 
-def run_ffmpeg_with_progress(commands : Commands, update_progress : UpdateProgress) -> subprocess.Popen[bytes]:
+def run_ffmpeg_with_progress(commands : List[Command], update_progress : UpdateProgress) -> subprocess.Popen[bytes]:
 	log_level = state_manager.get_item('log_level')
 	commands.extend(ffmpeg_builder.set_progress())
 	commands.extend(ffmpeg_builder.cast_stream())
@@ -22,8 +22,10 @@ def run_ffmpeg_with_progress(commands : Commands, update_progress : UpdateProgre
 
 	while process_manager.is_processing():
 		try:
-
 			while __line__ := process.stdout.readline().decode().lower():
+				if process_manager.is_stopping():
+					process.terminate()
+
 				if 'frame=' in __line__:
 					_, frame_number = __line__.split('frame=')
 					update_progress(int(frame_number))
@@ -35,12 +37,14 @@ def run_ffmpeg_with_progress(commands : Commands, update_progress : UpdateProgre
 			continue
 		return process
 
-	if process_manager.is_stopping():
-		process.terminate()
 	return process
 
 
-def run_ffmpeg(commands : Commands) -> subprocess.Popen[bytes]:
+def update_progress(progress : tqdm, frame_number : int) -> None:
+	progress.update(frame_number - progress.n)
+
+
+def run_ffmpeg(commands : List[Command]) -> subprocess.Popen[bytes]:
 	log_level = state_manager.get_item('log_level')
 	commands = ffmpeg_builder.run(commands)
 	process = subprocess.Popen(commands, stderr = subprocess.PIPE, stdout = subprocess.PIPE)
@@ -56,14 +60,60 @@ def run_ffmpeg(commands : Commands) -> subprocess.Popen[bytes]:
 
 	if process_manager.is_stopping():
 		process.terminate()
+
 	return process
 
 
-def open_ffmpeg(commands : Commands) -> subprocess.Popen[bytes]:
+def open_ffmpeg(commands : List[Command]) -> subprocess.Popen[bytes]:
 	commands = ffmpeg_builder.run(commands)
 	return subprocess.Popen(commands, stdin = subprocess.PIPE, stdout = subprocess.PIPE)
 
 
+def create_video_reader(video_path : str, frame_number : int, video_metadata : VideoReaderMetadata) -> subprocess.Popen[bytes]:
+	commands = ffmpeg_builder.chain(
+		ffmpeg_builder.seek_to(frame_number / video_metadata.get('fps')),
+		ffmpeg_builder.set_input(video_path),
+		ffmpeg_builder.restrict_color_transfer(video_metadata.get('color_transfer')),
+		ffmpeg_builder.prevent_frame_drop(),
+		ffmpeg_builder.enforce_pixel_format('bgr24'),
+		ffmpeg_builder.set_output_format('rawvideo'),
+		ffmpeg_builder.cast_stream()
+	)
+
+	return open_ffmpeg(commands)
+
+
+def create_video_writer(target_path : str, temp_video_fps : Fps, temp_video_resolution : Resolution, output_video_resolution : Resolution, output_video_fps : Fps) -> subprocess.Popen[bytes]:
+	output_video_encoder = state_manager.get_item('output_video_encoder')
+	output_video_quality = state_manager.get_item('output_video_quality')
+	output_video_preset = state_manager.get_item('output_video_preset')
+	temp_video_path = get_temp_file_path(target_path)
+	temp_video_format = cast(VideoFormat, get_file_format(temp_video_path))
+	output_video_encoder = fix_video_encoder(temp_video_format, output_video_encoder)
+
+	commands = ffmpeg_builder.chain(
+		ffmpeg_builder.set_output_format('rawvideo'),
+		ffmpeg_builder.enforce_pixel_format(state_manager.get_item('temp_pixel_format')),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(temp_video_resolution)),
+		ffmpeg_builder.set_input_fps(temp_video_fps),
+		ffmpeg_builder.set_input('pipe:0'),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(output_video_resolution)),
+		ffmpeg_builder.set_video_encoder(output_video_encoder),
+		ffmpeg_builder.set_thread_count(16),
+		ffmpeg_builder.set_video_tag(output_video_encoder, temp_video_format),
+		ffmpeg_builder.set_video_quality(output_video_encoder, output_video_quality),
+		ffmpeg_builder.set_video_preset(output_video_encoder, output_video_preset),
+		ffmpeg_builder.concat(
+			ffmpeg_builder.set_video_fps(output_video_fps),
+			ffmpeg_builder.convert_color_space('bt709')
+		),
+		ffmpeg_builder.set_pixel_format(output_video_encoder),
+		ffmpeg_builder.force_output(temp_video_path)
+	)
+
+	return open_ffmpeg(commands)
+
+
 def log_debug(process : subprocess.Popen[bytes]) -> None:
 	_, stderr = process.communicate()
 	errors = stderr.decode().split(os.linesep)
@@ -101,43 +151,54 @@ def get_available_encoder_set() -> EncoderSet:
 	return available_encoder_set
 
 
-def extract_frames(target_path : str, temp_video_resolution : str, temp_video_fps : Fps, trim_frame_start : int, trim_frame_end : int) -> bool:
-	extract_frame_total = predict_video_frame_total(target_path, temp_video_fps, trim_frame_start, trim_frame_end)
-	temp_frames_pattern = get_temp_frames_pattern(target_path, '%08d')
+def extract_frames(target_path : str, temp_video_resolution : Resolution, temp_video_fps : Fps, trim_frame_start : int, trim_frame_end : int) -> bool:
+	color_transfer = ffprobe.extract_static_video_metadata(target_path).get('color_transfer')
+	extract_frame_total = vision.predict_video_frame_total(target_path, temp_video_fps, trim_frame_start, trim_frame_end)
+	temp_frame_pattern = get_temp_frame_pattern(target_path, '%08d')
+
 	commands = ffmpeg_builder.chain(
 		ffmpeg_builder.set_input(target_path),
-		ffmpeg_builder.set_media_resolution(temp_video_resolution),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(temp_video_resolution)),
 		ffmpeg_builder.set_frame_quality(0),
-		ffmpeg_builder.select_frame_range(trim_frame_start, trim_frame_end, temp_video_fps),
+		ffmpeg_builder.enforce_pixel_format('rgb24'),
+		ffmpeg_builder.concat(
+			ffmpeg_builder.select_frame_range(trim_frame_start, trim_frame_end, temp_video_fps),
+			ffmpeg_builder.restrict_color_transfer(color_transfer)
+		),
 		ffmpeg_builder.prevent_frame_drop(),
-		ffmpeg_builder.set_output(temp_frames_pattern)
+		ffmpeg_builder.set_start_number(trim_frame_start),
+		ffmpeg_builder.set_output(temp_frame_pattern)
 	)
 
-	with tqdm(total = extract_frame_total, desc = wording.get('extracting'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
-		process = run_ffmpeg_with_progress(commands, lambda frame_number: progress.update(frame_number - progress.n))
+	with tqdm(total = extract_frame_total, desc = translator.get('extracting'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
+		process = run_ffmpeg_with_progress(commands, partial(update_progress, progress))
 		return process.returncode == 0
 
 
-def copy_image(target_path : str, temp_image_resolution : str) -> bool:
-	temp_file_path = get_temp_file_path(target_path)
+def copy_image(target_path : str, temp_image_resolution : Resolution) -> bool:
+	temp_image_path = get_temp_file_path(target_path)
+
 	commands = ffmpeg_builder.chain(
 		ffmpeg_builder.set_input(target_path),
-		ffmpeg_builder.set_media_resolution(temp_image_resolution),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(temp_image_resolution)),
 		ffmpeg_builder.set_image_quality(target_path, 100),
-		ffmpeg_builder.force_output(temp_file_path)
+		ffmpeg_builder.force_output(temp_image_path)
 	)
+
 	return run_ffmpeg(commands).returncode == 0
 
 
-def finalize_image(target_path : str, output_path : str, output_image_resolution : str) -> bool:
+def finalize_image(target_path : str, output_path : str, output_image_resolution : Resolution) -> bool:
 	output_image_quality = state_manager.get_item('output_image_quality')
-	temp_file_path = get_temp_file_path(target_path)
+	temp_image_path = get_temp_file_path(target_path)
+
 	commands = ffmpeg_builder.chain(
-		ffmpeg_builder.set_input(temp_file_path),
-		ffmpeg_builder.set_media_resolution(output_image_resolution),
+		ffmpeg_builder.set_input(temp_image_path),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(output_image_resolution)),
 		ffmpeg_builder.set_image_quality(target_path, output_image_quality),
 		ffmpeg_builder.force_output(output_path)
 	)
+
 	return run_ffmpeg(commands).returncode == 0
 
 
@@ -162,12 +223,15 @@ def restore_audio(target_path : str, output_path : str, trim_frame_start : int,
 	output_audio_encoder = state_manager.get_item('output_audio_encoder')
 	output_audio_quality = state_manager.get_item('output_audio_quality')
 	output_audio_volume = state_manager.get_item('output_audio_volume')
-	target_video_fps = detect_video_fps(target_path)
-	temp_file_path = get_temp_file_path(target_path)
-	temp_video_duration = detect_video_duration(temp_file_path)
+	target_video_fps = vision.detect_video_fps(target_path)
+	temp_video_path = get_temp_file_path(target_path)
+	temp_video_format = cast(VideoFormat, get_file_format(temp_video_path))
+	temp_video_duration = vision.detect_video_duration(temp_video_path)
+	output_video_format = cast(VideoFormat, get_file_format(output_path))
+	output_audio_encoder = fix_audio_encoder(temp_video_format, output_audio_encoder)
 
 	commands = ffmpeg_builder.chain(
-		ffmpeg_builder.set_input(temp_file_path),
+		ffmpeg_builder.set_input(temp_video_path),
 		ffmpeg_builder.select_media_range(trim_frame_start, trim_frame_end, target_video_fps),
 		ffmpeg_builder.set_input(target_path),
 		ffmpeg_builder.copy_video_encoder(),
@@ -177,8 +241,10 @@ def restore_audio(target_path : str, output_path : str, trim_frame_start : int,
 		ffmpeg_builder.select_media_stream('0:v:0'),
 		ffmpeg_builder.select_media_stream('1:a:0'),
 		ffmpeg_builder.set_video_duration(temp_video_duration),
+		ffmpeg_builder.set_faststart(output_video_format),
 		ffmpeg_builder.force_output(output_path)
 	)
+
 	return run_ffmpeg(commands).returncode == 0
 
 
@@ -186,53 +252,63 @@ def replace_audio(target_path : str, audio_path : str, output_path : str) -> boo
 	output_audio_encoder = state_manager.get_item('output_audio_encoder')
 	output_audio_quality = state_manager.get_item('output_audio_quality')
 	output_audio_volume = state_manager.get_item('output_audio_volume')
-	temp_file_path = get_temp_file_path(target_path)
-	temp_video_duration = detect_video_duration(temp_file_path)
+	temp_video_path = get_temp_file_path(target_path)
+	temp_video_format = cast(VideoFormat, get_file_format(temp_video_path))
+	temp_video_duration = vision.detect_video_duration(temp_video_path)
+	output_video_format = cast(VideoFormat, get_file_format(output_path))
+	output_audio_encoder = fix_audio_encoder(temp_video_format, output_audio_encoder)
 
 	commands = ffmpeg_builder.chain(
-		ffmpeg_builder.set_input(temp_file_path),
+		ffmpeg_builder.set_input(temp_video_path),
 		ffmpeg_builder.set_input(audio_path),
 		ffmpeg_builder.copy_video_encoder(),
 		ffmpeg_builder.set_audio_encoder(output_audio_encoder),
 		ffmpeg_builder.set_audio_quality(output_audio_encoder, output_audio_quality),
 		ffmpeg_builder.set_audio_volume(output_audio_volume),
 		ffmpeg_builder.set_video_duration(temp_video_duration),
+		ffmpeg_builder.set_faststart(output_video_format),
 		ffmpeg_builder.force_output(output_path)
 	)
+
 	return run_ffmpeg(commands).returncode == 0
 
 
-def merge_video(target_path : str, temp_video_fps : Fps, output_video_resolution : str, output_video_fps : Fps, trim_frame_start : int, trim_frame_end : int) -> bool:
+def merge_video(target_path : str, temp_video_fps : Fps, output_video_resolution : Resolution, output_video_fps : Fps, trim_frame_start : int, trim_frame_end : int) -> bool:
 	output_video_encoder = state_manager.get_item('output_video_encoder')
 	output_video_quality = state_manager.get_item('output_video_quality')
 	output_video_preset = state_manager.get_item('output_video_preset')
-	merge_frame_total = predict_video_frame_total(target_path, output_video_fps, trim_frame_start, trim_frame_end)
-	temp_file_path = get_temp_file_path(target_path)
-	temp_frames_pattern = get_temp_frames_pattern(target_path, '%08d')
-
-	if get_file_format(target_path) == 'webm':
-		output_video_encoder = 'libvpx-vp9'
+	merge_frame_total = vision.predict_video_frame_total(target_path, output_video_fps, trim_frame_start, trim_frame_end)
+	temp_video_path = get_temp_file_path(target_path)
+	temp_video_format = cast(VideoFormat, get_file_format(temp_video_path))
+	temp_frame_pattern = get_temp_frame_pattern(target_path, '%08d')
+	output_video_encoder = fix_video_encoder(temp_video_format, output_video_encoder)
 
 	commands = ffmpeg_builder.chain(
-		ffmpeg_builder.set_conditional_fps(temp_video_fps),
-		ffmpeg_builder.set_input(temp_frames_pattern),
-		ffmpeg_builder.set_media_resolution(output_video_resolution),
+		ffmpeg_builder.set_input_fps(temp_video_fps),
+		ffmpeg_builder.set_start_number(trim_frame_start),
+		ffmpeg_builder.set_input(temp_frame_pattern),
+		ffmpeg_builder.set_media_resolution(vision.pack_resolution(output_video_resolution)),
 		ffmpeg_builder.set_video_encoder(output_video_encoder),
+		ffmpeg_builder.set_video_tag(output_video_encoder, temp_video_format),
 		ffmpeg_builder.set_video_quality(output_video_encoder, output_video_quality),
 		ffmpeg_builder.set_video_preset(output_video_encoder, output_video_preset),
-		ffmpeg_builder.set_video_fps(output_video_fps),
+		ffmpeg_builder.concat(
+			ffmpeg_builder.set_video_fps(output_video_fps),
+			ffmpeg_builder.keep_video_alpha(output_video_encoder),
+			ffmpeg_builder.convert_color_space('bt709')
+		),
 		ffmpeg_builder.set_pixel_format(output_video_encoder),
-		ffmpeg_builder.set_video_colorspace('bt709'),
-		ffmpeg_builder.force_output(temp_file_path)
+		ffmpeg_builder.force_output(temp_video_path)
 	)
 
-	with tqdm(total = merge_frame_total, desc = wording.get('merging'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
-		process = run_ffmpeg_with_progress(commands, lambda frame_number: progress.update(frame_number - progress.n))
+	with tqdm(total = merge_frame_total, desc = translator.get('merging'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
+		process = run_ffmpeg_with_progress(commands, partial(update_progress, progress))
 		return process.returncode == 0
 
 
 def concat_video(output_path : str, temp_output_paths : List[str]) -> bool:
-	concat_video_path = tempfile.mktemp()
+	file_descriptor, concat_video_path = tempfile.mkstemp()
+	os.close(file_descriptor)
 
 	with open(concat_video_path, 'w') as concat_video_file:
 		for temp_output_path in temp_output_paths:
@@ -241,14 +317,44 @@ def concat_video(output_path : str, temp_output_paths : List[str]) -> bool:
 		concat_video_file.close()
 
 	output_path = os.path.abspath(output_path)
+	output_video_format = cast(VideoFormat, get_file_format(output_path))
+
 	commands = ffmpeg_builder.chain(
 		ffmpeg_builder.unsafe_concat(),
 		ffmpeg_builder.set_input(concat_video_file.name),
 		ffmpeg_builder.copy_video_encoder(),
 		ffmpeg_builder.copy_audio_encoder(),
+		ffmpeg_builder.set_faststart(output_video_format),
 		ffmpeg_builder.force_output(output_path)
 	)
+
 	process = run_ffmpeg(commands)
 	process.communicate()
 	remove_file(concat_video_path)
 	return process.returncode == 0
+
+
+def fix_audio_encoder(video_format : VideoFormat, audio_encoder : AudioEncoder) -> AudioEncoder:
+	if video_format == 'avi' and audio_encoder == 'libopus':
+		return 'aac'
+	if video_format in [ 'm4v', 'mpeg', 'wmv' ]:
+		return 'aac'
+	if video_format == 'mov' and audio_encoder in [ 'flac', 'libopus' ]:
+		return 'aac'
+	if video_format == 'mxf':
+		return 'pcm_s16le'
+	if video_format == 'webm':
+		return 'libopus'
+	return audio_encoder
+
+
+def fix_video_encoder(video_format : VideoFormat, video_encoder : VideoEncoder) -> VideoEncoder:
+	if video_format in [ 'm4v', 'mpeg', 'mxf', 'wmv' ]:
+		return 'libx264'
+	if video_format in [ 'mkv', 'mp4' ] and video_encoder == 'rawvideo':
+		return 'libx264'
+	if video_format == 'mov' and video_encoder == 'libvpx-vp9':
+		return 'libx264'
+	if video_format == 'webm':
+		return 'libvpx-vp9'
+	return video_encoder

facefusion/ffmpeg_builder.py

@@ -1,50 +1,73 @@
 import itertools
 import shutil
-from typing import Optional
+from typing import List, Optional
 
 import numpy
 
 from facefusion.filesystem import get_file_format
-from facefusion.types import AudioEncoder, Commands, Duration, Fps, StreamMode, VideoEncoder, VideoPreset
+from facefusion.types import AudioEncoder, ColorSpace, ColorTransfer, Command, CommandSet, Duration, Fps, StreamMode, VideoEncoder, VideoFormat, VideoPreset
 
 
-def run(commands : Commands) -> Commands:
+def run(commands : List[Command]) -> List[Command]:
 	return [ shutil.which('ffmpeg'), '-loglevel', 'error' ] + commands
 
 
-def chain(*commands : Commands) -> Commands:
+def chain(*commands : List[Command]) -> List[Command]:
 	return list(itertools.chain(*commands))
 
 
-def get_encoders() -> Commands:
+def concat(*__commands__ : List[Command]) -> List[Command]:
+	commands = []
+	command_set : CommandSet = {}
+
+	for command in __commands__:
+		for argument, value in zip(command[::2], command[1::2]):
+			command_set.setdefault(argument, []).append(value)
+
+	for argument, values in command_set.items():
+		commands.append(argument)
+		commands.append(','.join(values))
+
+	return commands
+
+
+def get_encoders() -> List[Command]:
 	return [ '-encoders' ]
 
 
-def set_progress() -> Commands:
+def set_hardware_accelerator(value : str) -> List[Command]:
+	return [ '-hwaccel', value ]
+
+
+def set_progress() -> List[Command]:
 	return [ '-progress' ]
 
 
-def set_input(input_path : str) -> Commands:
+def set_input(input_path : str) -> List[Command]:
 	return [ '-i', input_path ]
 
 
-def set_conditional_fps(conditional_fps : Fps) -> Commands:
-	return [ '-r', str(conditional_fps) ]
+def set_input_fps(input_fps : Fps) -> List[Command]:
+	return [ '-r', str(input_fps) ]
 
 
-def set_output(output_path : str) -> Commands:
+def set_start_number(frame_number : int) -> List[Command]:
+	return [ '-start_number', str(frame_number) ]
+
+
+def set_output(output_path : str) -> List[Command]:
 	return [ output_path ]
 
 
-def force_output(output_path : str) -> Commands:
+def force_output(output_path : str) -> List[Command]:
 	return [ '-y', output_path ]
 
 
-def cast_stream() -> Commands:
+def cast_stream() -> List[Command]:
 	return [ '-' ]
 
 
-def set_stream_mode(stream_mode : StreamMode) -> Commands:
+def set_stream_mode(stream_mode : StreamMode) -> List[Command]:
 	if stream_mode == 'udp':
 		return [ '-f', 'mpegts' ]
 	if stream_mode == 'v4l2':
@@ -52,21 +75,39 @@ def set_stream_mode(stream_mode : StreamMode) -> Commands:
 	return []
 
 
-def unsafe_concat() -> Commands:
+def set_stream_quality(stream_quality : int) -> List[Command]:
+	return [ '-b:v', str(stream_quality) + 'k' ]
+
+
+def unsafe_concat() -> List[Command]:
 	return [ '-f', 'concat', '-safe', '0' ]
 
 
-def set_pixel_format(video_encoder : VideoEncoder) -> Commands:
+def seek_to(time : float) -> List[Command]:
+	return [ '-ss', str(time)]
+
+
+def set_output_format(output_format : str) -> List[Command]:
+	return [ '-f', output_format ]
+
+
+def enforce_pixel_format(pixel_format : str) -> List[Command]:
+	return [ '-pix_fmt', pixel_format ]
+
+
+def set_pixel_format(video_encoder : VideoEncoder) -> List[Command]:
 	if video_encoder == 'rawvideo':
 		return [ '-pix_fmt', 'rgb24' ]
+	if video_encoder == 'libvpx-vp9':
+		return [ '-pix_fmt', 'yuva420p' ]
 	return [ '-pix_fmt', 'yuv420p' ]
 
 
-def set_frame_quality(frame_quality : int) -> Commands:
+def set_frame_quality(frame_quality : int) -> List[Command]:
 	return [ '-q:v', str(frame_quality) ]
 
 
-def select_frame_range(frame_start : int, frame_end : int, video_fps : Fps) -> Commands:
+def select_frame_range(frame_start : int, frame_end : int, video_fps : Fps) -> List[Command]:
 	if isinstance(frame_start, int) and isinstance(frame_end, int):
 		return [ '-vf', 'trim=start_frame=' + str(frame_start) + ':end_frame=' + str(frame_end) + ',fps=' + str(video_fps) ]
 	if isinstance(frame_start, int):
@@ -76,11 +117,21 @@ def select_frame_range(frame_start : int, frame_end : int, video_fps : Fps) -> C
 	return [ '-vf', 'fps=' + str(video_fps) ]
 
 
-def prevent_frame_drop() -> Commands:
+def prevent_frame_drop() -> List[Command]:
 	return [ '-vsync', '0' ]
 
 
-def select_media_range(frame_start : int, frame_end : int, media_fps : Fps) -> Commands:
+def restrict_color_transfer(color_transfer : ColorTransfer) -> List[Command]:
+	if color_transfer in [ 'smpte2084', 'arib-std-b67' ]:
+		return [ '-vf', 'scale=out_primaries=bt709:out_transfer=bt709:intent=perceptual' ]
+	return []
+
+
+def convert_color_space(color_space : ColorSpace) -> List[Command]:
+	return [ '-vf', 'scale=out_color_matrix=' + color_space + ':out_range=tv,setparams=colorspace=' + color_space + ':color_primaries=' + color_space + ':color_trc=' + color_space ]
+
+
+def select_media_range(frame_start : int, frame_end : int, media_fps : Fps) -> List[Command]:
 	commands = []
 
 	if isinstance(frame_start, int):
@@ -90,35 +141,35 @@ def select_media_range(frame_start : int, frame_end : int, media_fps : Fps) -> C
 	return commands
 
 
-def select_media_stream(media_stream : str) -> Commands:
+def select_media_stream(media_stream : str) -> List[Command]:
 	return [ '-map', media_stream ]
 
 
-def set_media_resolution(video_resolution : str) -> Commands:
+def set_media_resolution(video_resolution : str) -> List[Command]:
 	return [ '-s', video_resolution ]
 
 
-def set_image_quality(image_path : str, image_quality : int) -> Commands:
+def set_image_quality(image_path : str, image_quality : int) -> List[Command]:
 	if get_file_format(image_path) == 'webp':
-		image_compression = image_quality
-	else:
-		image_compression = round(31 - (image_quality * 0.31))
+		return [ '-q:v', str(image_quality) ]
+
+	image_compression = round(31 - (image_quality * 0.31))
 	return [ '-q:v', str(image_compression) ]
 
 
-def set_audio_encoder(audio_codec : str) -> Commands:
+def set_audio_encoder(audio_codec : str) -> List[Command]:
 	return [ '-c:a', audio_codec ]
 
 
-def copy_audio_encoder() -> Commands:
+def copy_audio_encoder() -> List[Command]:
 	return set_audio_encoder('copy')
 
 
-def set_audio_sample_rate(audio_sample_rate : int) -> Commands:
+def set_audio_sample_rate(audio_sample_rate : int) -> List[Command]:
 	return [ '-ar', str(audio_sample_rate) ]
 
 
-def set_audio_sample_size(audio_sample_size : int) -> Commands:
+def set_audio_sample_size(audio_sample_size : int) -> List[Command]:
 	if audio_sample_size == 16:
 		return [ '-f', 's16le' ]
 	if audio_sample_size == 32:
@@ -126,62 +177,78 @@ def set_audio_sample_size(audio_sample_size : int) -> Commands:
 	return []
 
 
-def set_audio_channel_total(audio_channel_total : int) -> Commands:
+def set_audio_channel_total(audio_channel_total : int) -> List[Command]:
 	return [ '-ac', str(audio_channel_total) ]
 
 
-def set_audio_quality(audio_encoder : AudioEncoder, audio_quality : int) -> Commands:
+def set_audio_quality(audio_encoder : AudioEncoder, audio_quality : int) -> List[Command]:
 	if audio_encoder == 'aac':
-		audio_compression = round(numpy.interp(audio_quality, [ 0, 100 ], [ 0.1, 2.0 ]), 1)
+		audio_compression = numpy.round(numpy.interp(audio_quality, [ 0, 100 ], [ 0.1, 2.0 ]), 1).astype(float).item()
 		return [ '-q:a', str(audio_compression) ]
 	if audio_encoder == 'libmp3lame':
-		audio_compression = round(numpy.interp(audio_quality, [ 0, 100 ], [ 9, 0 ]))
+		audio_compression = numpy.round(numpy.interp(audio_quality, [ 0, 100 ], [ 9, 0 ])).astype(int).item()
 		return [ '-q:a', str(audio_compression) ]
 	if audio_encoder == 'libopus':
-		audio_bit_rate = round(numpy.interp(audio_quality, [ 0, 100 ], [ 64, 320 ]))
+		audio_bit_rate = numpy.round(numpy.interp(audio_quality, [ 0, 100 ], [ 64, 256 ])).astype(int).item()
 		return [ '-b:a', str(audio_bit_rate) + 'k' ]
 	if audio_encoder == 'libvorbis':
-		audio_compression = round(numpy.interp(audio_quality, [ 0, 100 ], [ -1, 10 ]), 1)
+		audio_compression = numpy.round(numpy.interp(audio_quality, [ 0, 100 ], [ -1, 10 ]), 1).astype(float).item()
 		return [ '-q:a', str(audio_compression) ]
 	return []
 
 
-def set_audio_volume(audio_volume : int) -> Commands:
+def set_audio_volume(audio_volume : int) -> List[Command]:
 	return [ '-filter:a', 'volume=' + str(audio_volume / 100) ]
 
 
-def set_video_encoder(video_encoder : str) -> Commands:
+def set_thread_count(thread_count : int) -> List[Command]:
+	return [ '-threads', str(thread_count) ]
+
+
+def set_video_encoder(video_encoder : str) -> List[Command]:
 	return [ '-c:v', video_encoder ]
 
 
-def copy_video_encoder() -> Commands:
+def copy_video_encoder() -> List[Command]:
 	return set_video_encoder('copy')
 
 
-def set_video_quality(video_encoder : VideoEncoder, video_quality : int) -> Commands:
-	if video_encoder in [ 'libx264', 'libx265' ]:
-		video_compression = round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ]))
+def set_faststart(video_format : VideoFormat) -> List[Command]:
+	if video_format in [ 'm4v', 'mov', 'mp4' ]:
+		return [ '-movflags', '+faststart' ]
+	return []
+
+
+def set_video_tag(video_encoder : VideoEncoder, video_format : VideoFormat) -> List[Command]:
+	if video_format in [ 'm4v', 'mov', 'mp4' ] and video_encoder in [ 'libx265', 'hevc_nvenc', 'hevc_amf', 'hevc_qsv', 'hevc_videotoolbox' ]:
+		return [ '-tag:v', 'hvc1' ]
+	return []
+
+
+def set_video_quality(video_encoder : VideoEncoder, video_quality : int) -> List[Command]:
+	if video_encoder in [ 'libx264', 'libx264rgb', 'libx265' ]:
+		video_compression = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ])).astype(int).item()
 		return [ '-crf', str(video_compression) ]
 	if video_encoder == 'libvpx-vp9':
-		video_compression = round(numpy.interp(video_quality, [ 0, 100 ], [ 63, 0 ]))
+		video_compression = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 63, 0 ])).astype(int).item()
 		return [ '-crf', str(video_compression) ]
 	if video_encoder in [ 'h264_nvenc', 'hevc_nvenc' ]:
-		video_compression = round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ]))
+		video_compression = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ])).astype(int).item()
 		return [ '-cq', str(video_compression) ]
 	if video_encoder in [ 'h264_amf', 'hevc_amf' ]:
-		video_compression = round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ]))
+		video_compression = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ])).astype(int).item()
 		return [ '-qp_i', str(video_compression), '-qp_p', str(video_compression), '-qp_b', str(video_compression) ]
 	if video_encoder in [ 'h264_qsv', 'hevc_qsv' ]:
-		video_compression = round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ]))
+		video_compression = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 51, 0 ])).astype(int).item()
 		return [ '-qp', str(video_compression) ]
 	if video_encoder in [ 'h264_videotoolbox', 'hevc_videotoolbox' ]:
-		video_bit_rate = round(numpy.interp(video_quality, [ 0, 100 ], [ 1024, 50512 ]))
+		video_bit_rate = numpy.round(numpy.interp(video_quality, [ 0, 100 ], [ 1024, 50512 ])).astype(int).item()
 		return [ '-b:v', str(video_bit_rate) + 'k' ]
 	return []
 
 
-def set_video_preset(video_encoder : VideoEncoder, video_preset : VideoPreset) -> Commands:
-	if video_encoder in [ 'libx264', 'libx265' ]:
+def set_video_preset(video_encoder : VideoEncoder, video_preset : VideoPreset) -> List[Command]:
+	if video_encoder in [ 'libx264', 'libx264rgb', 'libx265' ]:
 		return [ '-preset', video_preset ]
 	if video_encoder in [ 'h264_nvenc', 'hevc_nvenc' ]:
 		return [ '-preset', map_nvenc_preset(video_preset) ]
@@ -192,23 +259,25 @@ def set_video_preset(video_encoder : VideoEncoder, video_preset : VideoPreset) -
 	return []
 
 
-def set_video_colorspace(video_colorspace : str) -> Commands:
-	return [ '-colorspace', video_colorspace ]
+def set_video_fps(video_fps : Fps) -> List[Command]:
+	return [ '-vf', 'fps=' + str(video_fps) ]
 
 
-def set_video_fps(video_fps : Fps) -> Commands:
-	return [ '-vf', 'framerate=fps=' + str(video_fps) ]
-
-
-def set_video_duration(video_duration : Duration) -> Commands:
+def set_video_duration(video_duration : Duration) -> List[Command]:
 	return [ '-t', str(video_duration) ]
 
 
-def capture_video() -> Commands:
-	return [ '-f', 'rawvideo' ]
+def keep_video_alpha(video_encoder : VideoEncoder) -> List[Command]:
+	if video_encoder == 'libvpx-vp9':
+		return [ '-vf', 'format=yuva420p' ]
+	return []
 
 
-def ignore_video_stream() -> Commands:
+def capture_video() -> List[Command]:
+	return [ '-f', 'rawvideo', '-pix_fmt', 'rgb24' ]
+
+
+def ignore_video_stream() -> List[Command]:
 	return [ '-vn' ]
 
 

facefusion/ffprobe.py

@@ -0,0 +1,130 @@
+import subprocess
+from functools import lru_cache
+from typing import Dict, List
+
+from facefusion import ffprobe_builder
+from facefusion.types import AudioMetadata, Buffer, Command, Fps, VideoMetadata
+
+
+def run_ffprobe(commands : List[Command]) -> subprocess.Popen[Buffer]:
+	commands = ffprobe_builder.run(commands)
+	return subprocess.Popen(commands, stderr = subprocess.PIPE, stdout = subprocess.PIPE)
+
+
+def parse_entries(output : Buffer) -> Dict[str, str]:
+	media_entries = {}
+
+	if output:
+		lines = output.decode().strip().splitlines()
+
+		for line in lines:
+			if '=' in line:
+				key, value = line.split('=', 1)
+				media_entries[key] = value
+
+	return media_entries
+
+
+def probe_audio_entries(audio_path : str, entries : List[str]) -> Dict[str, str]:
+	commands = ffprobe_builder.chain(
+		ffprobe_builder.select_stream('a:0'),
+		ffprobe_builder.show_stream_entries(entries),
+		ffprobe_builder.format_to_key_value(),
+		ffprobe_builder.set_input(audio_path)
+	)
+
+	output, _ = run_ffprobe(commands).communicate()
+
+	return parse_entries(output)
+
+
+def probe_video_entries(video_path : str, entries : List[str]) -> Dict[str, str]:
+	commands = ffprobe_builder.chain(
+		ffprobe_builder.select_stream('v:0'),
+		ffprobe_builder.show_stream_entries(entries),
+		ffprobe_builder.format_to_key_value(),
+		ffprobe_builder.set_input(video_path)
+	)
+
+	output, _ = run_ffprobe(commands).communicate()
+
+	return parse_entries(output)
+
+
+def probe_format_entries(media_path : str, entries : List[str]) -> Dict[str, str]:
+	commands = ffprobe_builder.chain(
+		ffprobe_builder.show_format_entries(entries),
+		ffprobe_builder.format_to_key_value(),
+		ffprobe_builder.set_input(media_path)
+	)
+
+	output, _ = run_ffprobe(commands).communicate()
+
+	return parse_entries(output)
+
+
+@lru_cache(maxsize = 128)
+def extract_static_audio_metadata(audio_path : str) -> AudioMetadata:
+	return extract_audio_metadata(audio_path)
+
+
+def extract_audio_metadata(audio_path : str) -> AudioMetadata:
+	audio_entries = probe_audio_entries(audio_path, [ 'sample_rate', 'channels' ])
+	format_entries = probe_format_entries(audio_path, [ 'duration', 'bit_rate' ])
+
+	duration = float(format_entries.get('duration'))
+	sample_rate = int(audio_entries.get('sample_rate'))
+	frame_total = round(duration * sample_rate)
+	channel_total = int(audio_entries.get('channels'))
+	bit_rate = int(format_entries.get('bit_rate'))
+
+	audio_metadata : AudioMetadata =\
+	{
+		'duration' : duration,
+		'frame_total' : frame_total,
+		'channel_total' : channel_total,
+		'sample_rate' : sample_rate,
+		'bit_rate' : bit_rate
+	}
+
+	return audio_metadata
+
+
+@lru_cache(maxsize = 128)
+def extract_static_video_metadata(video_path : str) -> VideoMetadata:
+	return extract_video_metadata(video_path)
+
+
+def extract_video_metadata(video_path : str) -> VideoMetadata:
+	video_entries = probe_video_entries(video_path, [ 'width', 'height', 'r_frame_rate', 'color_transfer' ])
+	format_entries = probe_format_entries(video_path, [ 'duration', 'bit_rate' ])
+
+	duration = float(format_entries.get('duration'))
+	fps = extract_video_fps(video_entries.get('r_frame_rate'))
+	frame_total = round(duration * fps)
+	width = int(video_entries.get('width'))
+	height = int(video_entries.get('height'))
+	bit_rate = int(format_entries.get('bit_rate'))
+	color_transfer = video_entries.get('color_transfer', 'unknown')
+
+	video_metadata : VideoMetadata =\
+	{
+		'duration' : duration,
+		'frame_total' : frame_total,
+		'fps' : fps,
+		'resolution' : (width, height),
+		'bit_rate' : bit_rate,
+		'color_transfer' : color_transfer
+	}
+
+	return video_metadata
+
+
+def extract_video_fps(frame_rate : str) -> Fps:
+	if frame_rate and '/' in frame_rate:
+		numerator, denominator = frame_rate.split('/')
+
+		if int(numerator) and int(denominator):
+			return int(numerator) / int(denominator)
+
+	return 0.0

facefusion/ffprobe_builder.py

@@ -0,0 +1,33 @@
+import itertools
+import shutil
+from typing import List
+
+from facefusion.types import Command
+
+
+def run(commands : List[Command]) -> List[Command]:
+	return [ shutil.which('ffprobe'), '-loglevel', 'error' ] + commands
+
+
+def chain(*commands : List[Command]) -> List[Command]:
+	return list(itertools.chain(*commands))
+
+
+def select_stream(stream : str) -> List[Command]:
+	return [ '-select_streams', stream ]
+
+
+def show_stream_entries(entries : List[str]) -> List[Command]:
+	return [ '-show_entries', 'stream=' + ','.join(entries) ]
+
+
+def show_format_entries(entries : List[str]) -> List[Command]:
+	return [ '-show_entries', 'format=' + ','.join(entries) ]
+
+
+def format_to_key_value() -> List[Command]:
+	return [ '-of', 'default=noprint_wrappers=1' ]
+
+
+def set_input(input_path : str) -> List[Command]:
+	return [ '-i', input_path ]

facefusion/filesystem.py

@@ -36,6 +36,8 @@ def get_file_format(file_path : str) -> Optional[str]:
 			return 'jpeg'
 		if file_extension == '.tif':
 			return 'tiff'
+		if file_extension == '.mpg':
+			return 'mpeg'
 		return file_extension.lstrip('.')
 	return None
 
@@ -99,7 +101,7 @@ def has_video(video_paths : List[str]) -> bool:
 
 def are_videos(video_paths : List[str]) -> bool:
 	if video_paths:
-		return any(map(is_video, video_paths))
+		return all(map(is_video, video_paths))
 	return False
 
 

facefusion/frame_store.py

@@ -0,0 +1,35 @@
+from facefusion.types import FrameStoreSet, VisionFrame, VisionFrameSet
+
+FRAME_STORE_SET : FrameStoreSet = {}
+
+
+def get_frame_store(id : str) -> VisionFrameSet:
+	if id not in FRAME_STORE_SET:
+		FRAME_STORE_SET[id] = {}
+
+	return FRAME_STORE_SET.get(id)
+
+
+def set_frame(id : str, frame_number : int, vision_frame : VisionFrame) -> None:
+	frame_store = get_frame_store(id)
+	frame_store[frame_number] = vision_frame
+
+
+def select_frame_set(id : str, frame_start : int, frame_end : int) -> VisionFrameSet:
+	frame_store = get_frame_store(id)
+	frame_set = {}
+
+	for frame_number in range(frame_start, frame_end + 1):
+		if frame_number in frame_store:
+			frame_set[frame_number] = frame_store.get(frame_number)
+
+	return frame_set
+
+
+def reduce_frames(id : str, frame_min : int, frame_max : int) -> None:
+	FRAME_STORE_SET[id] = select_frame_set(id, frame_min, frame_max)
+
+
+def clear_frames(id : str) -> None:
+	if id in FRAME_STORE_SET:
+		del FRAME_STORE_SET[id]

facefusion/inference_manager.py

@@ -1,14 +1,19 @@
 import importlib
-from time import sleep
+import random
+from functools import lru_cache
+from time import sleep, time
 from typing import List
 
 from onnxruntime import InferenceSession
 
-from facefusion import process_manager, state_manager
+from facefusion import logger, process_manager, state_manager, translator
 from facefusion.app_context import detect_app_context
-from facefusion.execution import create_inference_session_providers
-from facefusion.filesystem import is_file
-from facefusion.types import DownloadSet, ExecutionProvider, InferencePool, InferencePoolSet
+from facefusion.common_helper import is_windows
+from facefusion.execution import create_inference_providers, has_execution_provider
+from facefusion.exit_helper import fatal_exit
+from facefusion.filesystem import get_file_name, is_file
+from facefusion.time_helper import calculate_end_time
+from facefusion.types import DownloadSet, ExecutionProvider, InferencePool, InferencePoolSet, InferenceProvider
 
 INFERENCE_POOL_SET : InferencePoolSet =\
 {
@@ -20,55 +25,91 @@ INFERENCE_POOL_SET : InferencePoolSet =\
 def get_inference_pool(module_name : str, model_names : List[str], model_source_set : DownloadSet) -> InferencePool:
 	while process_manager.is_checking():
 		sleep(0.5)
-	execution_device_id = state_manager.get_item('execution_device_id')
-	execution_providers = resolve_execution_providers(module_name)
+	execution_device_ids = state_manager.get_item('execution_device_ids')
+	execution_providers = state_manager.get_item('execution_providers')
 	app_context = detect_app_context()
-	inference_context = get_inference_context(module_name, model_names, execution_device_id, execution_providers)
 
-	if app_context == 'cli' and INFERENCE_POOL_SET.get('ui').get(inference_context):
-		INFERENCE_POOL_SET['cli'][inference_context] = INFERENCE_POOL_SET.get('ui').get(inference_context)
-	if app_context == 'ui' and INFERENCE_POOL_SET.get('cli').get(inference_context):
-		INFERENCE_POOL_SET['ui'][inference_context] = INFERENCE_POOL_SET.get('cli').get(inference_context)
-	if not INFERENCE_POOL_SET.get(app_context).get(inference_context):
-		INFERENCE_POOL_SET[app_context][inference_context] = create_inference_pool(model_source_set, execution_device_id, execution_providers)
+	for execution_device_id in execution_device_ids:
+		inference_context = get_inference_context(module_name, model_names, execution_device_id, execution_providers)
 
-	return INFERENCE_POOL_SET.get(app_context).get(inference_context)
+		if app_context == 'cli' and INFERENCE_POOL_SET.get('ui').get(inference_context):
+			INFERENCE_POOL_SET['cli'][inference_context] = INFERENCE_POOL_SET.get('ui').get(inference_context)
+		if app_context == 'ui' and INFERENCE_POOL_SET.get('cli').get(inference_context):
+			INFERENCE_POOL_SET['ui'][inference_context] = INFERENCE_POOL_SET.get('cli').get(inference_context)
+		if not INFERENCE_POOL_SET.get(app_context).get(inference_context):
+			inference_providers = resolve_static_inference_providers(module_name, execution_device_id)
+			INFERENCE_POOL_SET[app_context][inference_context] = create_inference_pool(model_source_set, inference_providers)
+
+	current_inference_context = get_inference_context(module_name, model_names, random.choice(execution_device_ids), execution_providers)
+	return INFERENCE_POOL_SET.get(app_context).get(current_inference_context)
 
 
-def create_inference_pool(model_source_set : DownloadSet, execution_device_id : str, execution_providers : List[ExecutionProvider]) -> InferencePool:
+def create_inference_pool(model_source_set : DownloadSet, inference_providers : List[InferenceProvider]) -> InferencePool:
 	inference_pool : InferencePool = {}
 
 	for model_name in model_source_set.keys():
 		model_path = model_source_set.get(model_name).get('path')
 		if is_file(model_path):
-			inference_pool[model_name] = create_inference_session(model_path, execution_device_id, execution_providers)
+			inference_pool[model_name] = create_inference_session(model_path, inference_providers)
 
 	return inference_pool
 
 
 def clear_inference_pool(module_name : str, model_names : List[str]) -> None:
-	execution_device_id = state_manager.get_item('execution_device_id')
-	execution_providers = resolve_execution_providers(module_name)
+	execution_device_ids = state_manager.get_item('execution_device_ids')
+	execution_providers = state_manager.get_item('execution_providers')
 	app_context = detect_app_context()
-	inference_context = get_inference_context(module_name, model_names, execution_device_id, execution_providers)
 
-	if INFERENCE_POOL_SET.get(app_context).get(inference_context):
-		del INFERENCE_POOL_SET[app_context][inference_context]
+	if is_windows() and has_execution_provider('directml'):
+		INFERENCE_POOL_SET[app_context].clear()
+
+	for execution_device_id in execution_device_ids:
+		inference_context = get_inference_context(module_name, model_names, execution_device_id, execution_providers)
+		if INFERENCE_POOL_SET.get(app_context).get(inference_context):
+			del INFERENCE_POOL_SET[app_context][inference_context]
 
 
-def create_inference_session(model_path : str, execution_device_id : str, execution_providers : List[ExecutionProvider]) -> InferenceSession:
-	inference_session_providers = create_inference_session_providers(execution_device_id, execution_providers)
-	return InferenceSession(model_path, providers = inference_session_providers)
+def create_inference_session(model_path : str, inference_providers : List[InferenceProvider]) -> InferenceSession:
+	model_file_name = get_file_name(model_path)
+	start_time = time()
+
+	try:
+		inference_session = InferenceSession(model_path, providers = inference_providers)
+		logger.debug(translator.get('loading_model_succeeded').format(model_name = model_file_name, seconds = calculate_end_time(start_time)), __name__)
+		return inference_session
+
+	except Exception:
+		logger.error(translator.get('loading_model_failed').format(model_name = model_file_name), __name__)
+		fatal_exit(1)
 
 
-def get_inference_context(module_name : str, model_names : List[str], execution_device_id : str, execution_providers : List[ExecutionProvider]) -> str:
-	inference_context = '.'.join([ module_name ] + model_names + [ execution_device_id ] + list(execution_providers))
+def get_inference_context(module_name : str, model_names : List[str], execution_device_id : int, execution_providers : List[ExecutionProvider]) -> str:
+	inference_context = '.'.join([ module_name ] + model_names + [ str(execution_device_id) ] + list(execution_providers))
 	return inference_context
 
 
-def resolve_execution_providers(module_name : str) -> List[ExecutionProvider]:
+@lru_cache()
+def resolve_static_inference_providers(module_name : str, execution_device_id : int) -> List[InferenceProvider]:
 	module = importlib.import_module(module_name)
+	execution_providers = state_manager.get_item('execution_providers')
 
-	if hasattr(module, 'resolve_execution_providers'):
-		return getattr(module, 'resolve_execution_providers')()
-	return state_manager.get_item('execution_providers')
+	if hasattr(module, 'override_inference_providers'):
+		override_inference_providers = getattr(module, 'override_inference_providers')()
+
+		if override_inference_providers:
+			return override_inference_providers
+
+	if hasattr(module, 'adjust_inference_providers'):
+		adjust_inference_providers = getattr(module, 'adjust_inference_providers')()
+
+		if adjust_inference_providers:
+			inference_providers = create_inference_providers(execution_device_id, execution_providers)
+
+			for adjust_inference_provider in adjust_inference_providers:
+				for inference_provider in inference_providers:
+					if inference_provider[0] == adjust_inference_provider[0] and inference_provider[1]:
+						inference_provider[1].update(adjust_inference_provider[1])
+
+			return inference_providers
+
+	return create_inference_providers(execution_device_id, execution_providers)

facefusion/installer.py

@@ -4,87 +4,71 @@ import signal
 import subprocess
 import sys
 from argparse import ArgumentParser, HelpFormatter
+from functools import partial
+from types import FrameType
 
-from facefusion import metadata, wording
+from facefusion import metadata
 from facefusion.common_helper import is_linux, is_windows
 
+LOCALES =\
+{
+	'install_dependency': 'install the {dependency} package',
+	'force_reinstall': 'force reinstall of packages',
+	'skip_conda': 'skip the conda environment check',
+	'conda_not_activated': 'conda is not activated'
+}
 ONNXRUNTIME_SET =\
 {
-	'default': ('onnxruntime', '1.21.1')
+	'default': ('onnxruntime', '1.26.0')
 }
 if is_windows() or is_linux():
-	ONNXRUNTIME_SET['cuda'] = ('onnxruntime-gpu', '1.21.1')
-	ONNXRUNTIME_SET['openvino'] = ('onnxruntime-openvino', '1.21.0')
+	ONNXRUNTIME_SET['cuda'] = ('onnxruntime-gpu', '1.26.0')
+	ONNXRUNTIME_SET['openvino'] = ('onnxruntime-openvino', '1.24.1')
 if is_windows():
-	ONNXRUNTIME_SET['directml'] = ('onnxruntime-directml', '1.17.3')
+	ONNXRUNTIME_SET['directml'] = ('onnxruntime-directml', '1.24.4')
+	ONNXRUNTIME_SET['qnn'] = ('onnxruntime-qnn', '1.24.4')
 if is_linux():
-	ONNXRUNTIME_SET['rocm'] = ('onnxruntime-rocm', '1.21.0')
+	ONNXRUNTIME_SET['migraphx'] = ('onnxruntime-migraphx', '1.25.0')
+	ONNXRUNTIME_SET['rocm'] = ('onnxruntime-rocm', '1.22.2.post1')
 
 
 def cli() -> None:
-	signal.signal(signal.SIGINT, lambda signal_number, frame: sys.exit(0))
-	program = ArgumentParser(formatter_class = lambda prog: HelpFormatter(prog, max_help_position = 50))
-	program.add_argument('--onnxruntime', help = wording.get('help.install_dependency').format(dependency = 'onnxruntime'), choices = ONNXRUNTIME_SET.keys(), required = True)
-	program.add_argument('--skip-conda', help = wording.get('help.skip_conda'), action = 'store_true')
+	signal.signal(signal.SIGINT, signal_exit)
+	program = ArgumentParser(formatter_class = partial(HelpFormatter, max_help_position = 50))
+	program.add_argument('onnxruntime', help = LOCALES.get('install_dependency').format(dependency = 'onnxruntime'), choices = ONNXRUNTIME_SET.keys())
+	program.add_argument('--force-reinstall', help = LOCALES.get('force_reinstall'), action = 'store_true')
+	program.add_argument('--skip-conda', help = LOCALES.get('skip_conda'), action = 'store_true')
 	program.add_argument('-v', '--version', version = metadata.get('name') + ' ' + metadata.get('version'), action = 'version')
 	run(program)
 
 
+def signal_exit(signum : int, frame : FrameType) -> None:
+	sys.exit(0)
+
+
 def run(program : ArgumentParser) -> None:
 	args = program.parse_args()
 	has_conda = 'CONDA_PREFIX' in os.environ
-	onnxruntime_name, onnxruntime_version = ONNXRUNTIME_SET.get(args.onnxruntime)
 
 	if not args.skip_conda and not has_conda:
-		sys.stdout.write(wording.get('conda_not_activated') + os.linesep)
+		sys.stdout.write(LOCALES.get('conda_not_activated') + os.linesep)
 		sys.exit(1)
 
+	commands = [ shutil.which('pip'), 'install' ]
+
+	if args.force_reinstall:
+		commands.append('--force-reinstall')
+
 	with open('requirements.txt') as file:
 
 		for line in file.readlines():
 			__line__ = line.strip()
 			if not __line__.startswith('onnxruntime'):
-				subprocess.call([ shutil.which('pip'), 'install', line, '--force-reinstall' ])
+				commands.append(__line__)
 
-	if args.onnxruntime == 'rocm':
-		python_id = 'cp' + str(sys.version_info.major) + str(sys.version_info.minor)
+	onnxruntime_name, onnxruntime_version = ONNXRUNTIME_SET.get(args.onnxruntime)
+	commands.append(onnxruntime_name + '==' + onnxruntime_version)
 
-		if python_id in [ 'cp310', 'cp312' ]:
-			wheel_name = 'onnxruntime_rocm-' + onnxruntime_version + '-' + python_id + '-' + python_id + '-linux_x86_64.whl'
-			wheel_url = 'https://repo.radeon.com/rocm/manylinux/rocm-rel-6.4/' + wheel_name
-			subprocess.call([ shutil.which('pip'), 'install', wheel_url, '--force-reinstall' ])
-	else:
-		subprocess.call([ shutil.which('pip'), 'install', onnxruntime_name + '==' + onnxruntime_version, '--force-reinstall' ])
+	subprocess.call([ shutil.which('pip'), 'uninstall', 'onnxruntime', onnxruntime_name, '-y', '-q' ])
 
-	if args.onnxruntime == 'cuda' and has_conda:
-		library_paths = []
-
-		if is_linux():
-			if os.getenv('LD_LIBRARY_PATH'):
-				library_paths = os.getenv('LD_LIBRARY_PATH').split(os.pathsep)
-
-			python_id = 'python' + str(sys.version_info.major) + '.' + str(sys.version_info.minor)
-			library_paths.extend(
-			[
-				os.path.join(os.getenv('CONDA_PREFIX'), 'lib'),
-				os.path.join(os.getenv('CONDA_PREFIX'), 'lib', python_id, 'site-packages', 'tensorrt_libs')
-			])
-			library_paths = list(dict.fromkeys([ library_path for library_path in library_paths if os.path.exists(library_path) ]))
-
-			subprocess.call([ shutil.which('conda'), 'env', 'config', 'vars', 'set', 'LD_LIBRARY_PATH=' + os.pathsep.join(library_paths) ])
-
-		if is_windows():
-			if os.getenv('PATH'):
-				library_paths = os.getenv('PATH').split(os.pathsep)
-
-			library_paths.extend(
-			[
-				os.path.join(os.getenv('CONDA_PREFIX'), 'Lib'),
-				os.path.join(os.getenv('CONDA_PREFIX'), 'Lib', 'site-packages', 'tensorrt_libs')
-			])
-			library_paths = list(dict.fromkeys([ library_path for library_path in library_paths if os.path.exists(library_path) ]))
-
-			subprocess.call([ shutil.which('conda'), 'env', 'config', 'vars', 'set', 'PATH=' + os.pathsep.join(library_paths) ])
-
-	if args.onnxruntime == 'directml':
-		subprocess.call([ shutil.which('pip'), 'install', 'numpy==1.26.4', '--force-reinstall' ])
+	subprocess.call(commands)

facefusion/jobs/job_helper.py

@@ -7,10 +7,12 @@ from facefusion.filesystem import get_file_extension, get_file_name
 
 def get_step_output_path(job_id : str, step_index : int, output_path : str) -> Optional[str]:
 	if output_path:
-		output_directory_path, _ = os.path.split(output_path)
-		output_file_name = get_file_name(_)
-		output_file_extension = get_file_extension(_)
-		return os.path.join(output_directory_path, output_file_name + '-' + job_id + '-' + str(step_index) + output_file_extension)
+		output_directory_path, output_file_path = os.path.split(output_path)
+		output_file_name = get_file_name(output_file_path)
+		output_file_extension = get_file_extension(output_file_path)
+
+		if output_file_name and output_file_extension:
+			return os.path.join(output_directory_path, output_file_name + '-' + job_id + '-' + str(step_index) + output_file_extension)
 	return None
 
 

facefusion/jobs/job_list.py

@@ -1,15 +1,15 @@
 from datetime import datetime
-from typing import Optional, Tuple
+from typing import List, Optional, Tuple
 
-from facefusion.date_helper import describe_time_ago
 from facefusion.jobs import job_manager
-from facefusion.types import JobStatus, TableContents, TableHeaders
+from facefusion.time_helper import describe_time_ago
+from facefusion.types import JobStatus, TableContent, TableHeader
 
 
-def compose_job_list(job_status : JobStatus) -> Tuple[TableHeaders, TableContents]:
+def compose_job_list(job_status : JobStatus) -> Tuple[List[TableHeader], List[List[TableContent]]]:
 	jobs = job_manager.find_jobs(job_status)
-	job_headers : TableHeaders = [ 'job id', 'steps', 'date created', 'date updated', 'job status' ]
-	job_contents : TableContents = []
+	job_headers : List[TableHeader] = [ 'job id', 'steps', 'date created', 'date updated', 'job status' ]
+	job_contents : List[List[TableContent]] = []
 
 	for index, job_id in enumerate(jobs):
 		if job_manager.validate_job(job_id):

facefusion/jobs/job_manager.py

@@ -3,10 +3,11 @@ from copy import copy
 from typing import List, Optional
 
 import facefusion.choices
-from facefusion.date_helper import get_current_date_time
 from facefusion.filesystem import create_directory, get_file_name, is_directory, is_file, move_file, remove_directory, remove_file, resolve_file_pattern
 from facefusion.jobs.job_helper import get_step_output_path
 from facefusion.json import read_json, write_json
+from facefusion.sanitizer import sanitize_job_id
+from facefusion.time_helper import get_current_date_time
 from facefusion.types import Args, Job, JobSet, JobStatus, JobStep, JobStepStatus
 
 JOBS_PATH : Optional[str] = None
@@ -261,5 +262,6 @@ def find_job_path(job_id : str) -> Optional[str]:
 
 def get_job_file_name(job_id : str) -> Optional[str]:
 	if job_id:
+		job_id = sanitize_job_id(job_id)
 		return job_id + '.json'
 	return None

facefusion/jobs/job_store.py

@@ -17,11 +17,11 @@ def get_step_keys() -> List[str]:
 	return JOB_STORE.get('step_keys')
 
 
-def register_job_keys(step_keys : List[str]) -> None:
-	for step_key in step_keys:
-		JOB_STORE['job_keys'].append(step_key)
-
-
-def register_step_keys(job_keys : List[str]) -> None:
+def register_job_keys(job_keys : List[str]) -> None:
 	for job_key in job_keys:
-		JOB_STORE['step_keys'].append(job_key)
+		JOB_STORE['job_keys'].append(job_key)
+
+
+def register_step_keys(step_keys : List[str]) -> None:
+	for step_key in step_keys:
+		JOB_STORE['step_keys'].append(step_key)

facefusion/locales.py

@@ -0,0 +1,277 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'conda_not_activated': 'conda is not activated',
+		'python_not_supported': 'python version is not supported, upgrade to {version} or higher',
+		'dependency_not_installed': '{dependency} is not installed',
+		'creating_temp': 'creating temporary resources',
+		'extracting_frames': 'extracting frames with a resolution of {resolution} and {fps} frames per second',
+		'extracting_frames_succeeded': 'extracting frames succeeded',
+		'extracting_frames_failed': 'extracting frames failed',
+		'analysing': 'analysing',
+		'extracting': 'extracting',
+		'streaming': 'streaming',
+		'processing': 'processing',
+		'merging': 'merging',
+		'downloading': 'downloading',
+		'temp_frames_not_found': 'temporary frames not found',
+		'copying_image': 'copying image with a resolution of {resolution}',
+		'copying_image_succeeded': 'copying image succeeded',
+		'copying_image_failed': 'copying image failed',
+		'finalizing_image': 'finalizing image with a resolution of {resolution}',
+		'finalizing_image_succeeded': 'finalizing image succeeded',
+		'finalizing_image_skipped': 'finalizing image skipped',
+		'merging_video': 'merging video with a resolution of {resolution} and {fps} frames per second',
+		'merging_video_succeeded': 'merging video succeeded',
+		'merging_video_failed': 'merging video failed',
+		'skipping_audio': 'skipping audio',
+		'replacing_audio_succeeded': 'replacing audio succeeded',
+		'replacing_audio_skipped': 'replacing audio skipped',
+		'restoring_audio_succeeded': 'restoring audio succeeded',
+		'restoring_audio_skipped': 'restoring audio skipped',
+		'clearing_temp': 'clearing temporary resources',
+		'processing_stopped': 'processing stopped',
+		'processing_image_succeeded': 'processing to image succeeded in {seconds} seconds',
+		'processing_image_failed': 'processing to image failed',
+		'processing_video_succeeded': 'processing to video succeeded in {seconds} seconds',
+		'processing_video_failed': 'processing to video failed',
+		'choose_image_source': 'choose an image for the source',
+		'choose_audio_source': 'choose an audio for the source',
+		'choose_video_target': 'choose a video for the target',
+		'choose_image_or_video_target': 'choose an image or video for the target',
+		'specify_image_or_video_output': 'specify the output image or video within a directory',
+		'match_target_and_output_extension': 'match the target and output extension',
+		'no_source_face_detected': 'no source face detected',
+		'processor_not_loaded': 'processor {processor} could not be loaded',
+		'processor_not_implemented': 'processor {processor} not implemented correctly',
+		'ui_layout_not_loaded': 'ui layout {ui_layout} could not be loaded',
+		'ui_layout_not_implemented': 'ui layout {ui_layout} not implemented correctly',
+		'stream_not_loaded': 'stream {stream_mode} could not be loaded',
+		'stream_not_supported': 'stream not supported',
+		'job_created': 'job {job_id} created',
+		'job_not_created': 'job {job_id} not created',
+		'job_submitted': 'job {job_id} submitted',
+		'job_not_submitted': 'job {job_id} not submitted',
+		'job_all_submitted': 'jobs submitted',
+		'job_all_not_submitted': 'jobs not submitted',
+		'job_deleted': 'job {job_id} deleted',
+		'job_not_deleted': 'job {job_id} not deleted',
+		'job_all_deleted': 'jobs deleted',
+		'job_all_not_deleted': 'jobs not deleted',
+		'job_step_added': 'step added to job {job_id}',
+		'job_step_not_added': 'step not added to job {job_id}',
+		'job_remix_step_added': 'step {step_index} remixed from job {job_id}',
+		'job_remix_step_not_added': 'step {step_index} not remixed from job {job_id}',
+		'job_step_inserted': 'step {step_index} inserted to job {job_id}',
+		'job_step_not_inserted': 'step {step_index} not inserted to job {job_id}',
+		'job_step_removed': 'step {step_index} removed from job {job_id}',
+		'job_step_not_removed': 'step {step_index} not removed from job {job_id}',
+		'running_job': 'running queued job {job_id}',
+		'running_jobs': 'running all queued jobs',
+		'retrying_job': 'retrying failed job {job_id}',
+		'retrying_jobs': 'retrying all failed jobs',
+		'processing_job_succeeded': 'processing of job {job_id} succeeded',
+		'processing_jobs_succeeded': 'processing of all jobs succeeded',
+		'processing_job_failed': 'processing of job {job_id} failed',
+		'processing_jobs_failed': 'processing of all jobs failed',
+		'processing_step': 'processing step {step_current} of {step_total}',
+		'validating_hash_succeeded': 'validating hash for {hash_file_name} succeeded',
+		'validating_hash_failed': 'validating hash for {hash_file_name} failed',
+		'validating_source_succeeded': 'validating source for {source_file_name} succeeded',
+		'validating_source_failed': 'validating source for {source_file_name} failed',
+		'deleting_corrupt_source': 'deleting corrupt source for {source_file_name}',
+		'loading_model_succeeded': 'loading model {model_name} succeeded in {seconds} seconds',
+		'loading_model_failed': 'loading model {model_name} failed',
+		'time_ago_now': 'just now',
+		'time_ago_minutes': '{minutes} minutes ago',
+		'time_ago_hours': '{hours} hours and {minutes} minutes ago',
+		'time_ago_days': '{days} days, {hours} hours and {minutes} minutes ago',
+		'point': '.',
+		'comma': ',',
+		'colon': ':',
+		'question_mark': '?',
+		'exclamation_mark': '!',
+		'help':
+		{
+			'install_dependency': 'choose the variant of {dependency} to install',
+			'skip_conda': 'skip the conda environment check',
+			'workflow_mode': 'choose the workflow mode',
+			'workflow_strategy': 'choose the workflow strategy',
+			'config_path': 'choose the config file to override defaults',
+			'temp_path': 'specify the directory for the temporary resources',
+			'jobs_path': 'specify the directory to store jobs',
+			'source_paths': 'choose the image or audio paths',
+			'target_path': 'choose the image or video path',
+			'output_path': 'specify the image or video within a directory',
+			'source_pattern': 'choose the image or audio pattern',
+			'target_pattern': 'choose the image or video pattern',
+			'output_pattern': 'specify the image or video pattern',
+			'face_detector_model': 'choose the model responsible for detecting the faces',
+			'face_detector_size': 'specify the frame size provided to the face detector',
+			'face_detector_margin': 'apply top, right, bottom and left margin to the frame',
+			'face_detector_angles': 'specify the angles to rotate the frame before detecting faces',
+			'face_detector_score': 'filter the detected faces based on the confidence score',
+			'face_landmarker_model': 'choose the model responsible for detecting the face landmarks',
+			'face_landmarker_score': 'filter the detected face landmarks based on the confidence score',
+			'face_selector_mode': 'use reference based tracking or simple matching',
+			'face_selector_order': 'specify the order of the detected faces',
+			'face_selector_age_start': 'filter the detected faces based on the starting age',
+			'face_selector_age_end': 'filter the detected faces based on the ending age',
+			'face_selector_gender': 'filter the detected faces based on their gender',
+			'face_selector_race': 'filter the detected faces based on their race',
+			'reference_face_position': 'specify the position used to create the reference face',
+			'reference_face_distance': 'specify the similarity between the reference face and target face',
+			'reference_frame_number': 'specify the frame used to create the reference face',
+			'face_tracker_score': 'specify the overlap score used to match the tracked faces',
+			'face_occluder_model': 'choose the model responsible for the occlusion mask',
+			'face_parser_model': 'choose the model responsible for the region mask',
+			'face_mask_types': 'mix and match different face mask types (choices: {choices})',
+			'face_mask_areas': 'choose the items used for the area mask (choices: {choices})',
+			'face_mask_regions': 'choose the items used for the region mask (choices: {choices})',
+			'face_mask_blur': 'specify the degree of blur applied to the box mask',
+			'face_mask_padding': 'apply top, right, bottom and left padding to the box mask',
+			'voice_extractor_model': 'choose the model responsible for extracting the voices',
+			'trim_frame_start': 'specify the starting frame of the target video',
+			'trim_frame_end': 'specify the ending frame of the target video',
+			'temp_frame_format': 'specify the temporary resources format',
+			'temp_pixel_format': 'specify the temporary pixel format',
+			'target_frame_amount': 'specify the amount of target frames forwarded to the processor',
+			'output_image_quality': 'specify the image quality which translates to the image compression',
+			'output_image_scale': 'specify the image scale based on the target image',
+			'output_audio_encoder': 'specify the encoder used for the audio',
+			'output_audio_quality': 'specify the audio quality which translates to the audio compression',
+			'output_audio_volume': 'specify the audio volume based on the target video',
+			'output_video_encoder': 'specify the encoder used for the video',
+			'output_video_preset': 'balance fast video processing and video file size',
+			'output_video_quality': 'specify the video quality which translates to the video compression',
+			'output_video_scale': 'specify the video scale based on the target video',
+			'output_video_fps': 'specify the video fps based on the target video',
+			'processors': 'load a single or multiple processors (choices: {choices}, ...)',
+			'background-remover-model': 'choose the model responsible for removing the background',
+			'background-remover-color': 'apply red, green blue and alpha values of the background',
+			'open_browser': 'open the browser once the program is ready',
+			'ui_layouts': 'launch a single or multiple UI layouts (choices: {choices}, ...)',
+			'ui_workflow': 'choose the ui workflow',
+			'download_providers': 'download using different providers (choices: {choices}, ...)',
+			'download_scope': 'specify the download scope',
+			'benchmark_mode': 'choose the benchmark mode',
+			'benchmark_resolutions': 'choose the resolutions for the benchmarks (choices: {choices}, ...)',
+			'benchmark_cycle_count': 'specify the amount of cycles per benchmark',
+			'execution_device_ids': 'specify the devices used for processing',
+			'execution_providers': 'inference using different providers (choices: {choices}, ...)',
+			'execution_thread_count': 'specify the amount of parallel threads while processing',
+			'video_memory_strategy': 'balance fast processing and low VRAM usage',
+			'log_level': 'adjust the message severity displayed in the terminal',
+			'halt_on_error': 'halt the program once an error occurred',
+			'run': 'run the program',
+			'headless_run': 'run the program in headless mode',
+			'batch_run': 'run the program in batch mode',
+			'force_download': 'force automate downloads and exit',
+			'benchmark': 'benchmark the program',
+			'job_id': 'specify the job id',
+			'job_status': 'specify the job status',
+			'step_index': 'specify the step index',
+			'job_list': 'list jobs by status',
+			'job_create': 'create a drafted job',
+			'job_submit': 'submit a drafted job to become a queued job',
+			'job_submit_all': 'submit all drafted jobs to become a queued jobs',
+			'job_delete': 'delete a drafted, queued, failed or completed job',
+			'job_delete_all': 'delete all drafted, queued, failed and completed jobs',
+			'job_add_step': 'add a step to a drafted job',
+			'job_remix_step': 'remix a previous step from a drafted job',
+			'job_insert_step': 'insert a step to a drafted job',
+			'job_remove_step': 'remove a step from a drafted job',
+			'job_run': 'run a queued job',
+			'job_run_all': 'run all queued jobs',
+			'job_retry': 'retry a failed job',
+			'job_retry_all': 'retry all failed jobs'
+		},
+		'about':
+		{
+			'fund': 'fund ai workstation',
+			'subscribe': 'become a member',
+			'join': 'join our community'
+		},
+		'uis':
+		{
+			'apply_button': 'APPLY',
+			'benchmark_mode_dropdown': 'BENCHMARK MODE',
+			'benchmark_cycle_count_slider': 'BENCHMARK CYCLE COUNT',
+			'benchmark_resolutions_checkbox_group': 'BENCHMARK RESOLUTIONS',
+			'clear_button': 'CLEAR',
+			'download_providers_checkbox_group': 'DOWNLOAD PROVIDERS',
+			'execution_providers_checkbox_group': 'EXECUTION PROVIDERS',
+			'workflow_mode_dropdown': 'WORKFLOW MODE',
+			'workflow_strategy_dropdown': 'WORKFLOW STRATEGY',
+			'execution_thread_count_slider': 'EXECUTION THREAD COUNT',
+			'face_detector_angles_checkbox_group': 'FACE DETECTOR ANGLES',
+			'face_detector_model_dropdown': 'FACE DETECTOR MODEL',
+			'face_detector_margin_slider': 'FACE DETECTOR MARGIN',
+			'face_detector_score_slider': 'FACE DETECTOR SCORE',
+			'face_detector_size_dropdown': 'FACE DETECTOR SIZE',
+			'face_landmarker_model_dropdown': 'FACE LANDMARKER MODEL',
+			'face_landmarker_score_slider': 'FACE LANDMARKER SCORE',
+			'face_mask_blur_slider': 'FACE MASK BLUR',
+			'face_mask_padding_bottom_slider': 'FACE MASK PADDING BOTTOM',
+			'face_mask_padding_left_slider': 'FACE MASK PADDING LEFT',
+			'face_mask_padding_right_slider': 'FACE MASK PADDING RIGHT',
+			'face_mask_padding_top_slider': 'FACE MASK PADDING TOP',
+			'face_mask_areas_checkbox_group': 'FACE MASK AREAS',
+			'face_mask_regions_checkbox_group': 'FACE MASK REGIONS',
+			'face_mask_types_checkbox_group': 'FACE MASK TYPES',
+			'face_selector_age_range_slider': 'FACE SELECTOR AGE',
+			'face_selector_gender_dropdown': 'FACE SELECTOR GENDER',
+			'face_selector_mode_dropdown': 'FACE SELECTOR MODE',
+			'face_selector_order_dropdown': 'FACE SELECTOR ORDER',
+			'face_selector_race_dropdown': 'FACE SELECTOR RACE',
+			'face_tracker_score_slider': 'FACE TRACKER SCORE',
+			'face_occluder_model_dropdown': 'FACE OCCLUDER MODEL',
+			'face_parser_model_dropdown': 'FACE PARSER MODEL',
+			'voice_extractor_model_dropdown': 'VOICE EXTRACTOR MODEL',
+			'job_list_status_checkbox_group': 'JOB STATUS',
+			'job_manager_job_action_dropdown': 'JOB_ACTION',
+			'job_manager_job_id_dropdown': 'JOB ID',
+			'job_manager_step_index_dropdown': 'STEP INDEX',
+			'job_runner_job_action_dropdown': 'JOB ACTION',
+			'job_runner_job_id_dropdown': 'JOB ID',
+			'log_level_dropdown': 'LOG LEVEL',
+			'output_audio_encoder_dropdown': 'OUTPUT AUDIO ENCODER',
+			'output_audio_quality_slider': 'OUTPUT AUDIO QUALITY',
+			'output_audio_volume_slider': 'OUTPUT AUDIO VOLUME',
+			'output_image_or_video': 'OUTPUT',
+			'output_image_quality_slider': 'OUTPUT IMAGE QUALITY',
+			'output_image_scale_slider': 'OUTPUT IMAGE SCALE',
+			'output_path_textbox': 'OUTPUT PATH',
+			'output_video_encoder_dropdown': 'OUTPUT VIDEO ENCODER',
+			'output_video_fps_slider': 'OUTPUT VIDEO FPS',
+			'output_video_preset_dropdown': 'OUTPUT VIDEO PRESET',
+			'output_video_quality_slider': 'OUTPUT VIDEO QUALITY',
+			'output_video_scale_slider': 'OUTPUT VIDEO SCALE',
+			'preview_frame_slider': 'PREVIEW FRAME',
+			'preview_image': 'PREVIEW',
+			'preview_mode_dropdown': 'PREVIEW MODE',
+			'preview_resolution_dropdown': 'PREVIEW RESOLUTION',
+			'processors_checkbox_group': 'PROCESSORS',
+			'reference_face_distance_slider': 'REFERENCE FACE DISTANCE',
+			'reference_face_gallery': 'REFERENCE FACE',
+			'refresh_button': 'REFRESH',
+			'source_file': 'SOURCE',
+			'start_button': 'START',
+			'stop_button': 'STOP',
+			'target_file': 'TARGET',
+			'temp_frame_format_dropdown': 'TEMP FRAME FORMAT',
+			'terminal_textbox': 'TERMINAL',
+			'trim_frame_slider': 'TRIM FRAME',
+			'ui_workflow': 'UI WORKFLOW',
+			'video_memory_strategy_dropdown': 'VIDEO MEMORY STRATEGY',
+			'webcam_fps_slider': 'WEBCAM FPS',
+			'webcam_image': 'WEBCAM',
+			'webcam_device_id_dropdown': 'WEBCAM DEVICE ID',
+			'webcam_mode_radio': 'WEBCAM MODE',
+			'webcam_resolution_dropdown': 'WEBCAM RESOLUTION'
+		}
+	}
+}

facefusion/memory.py

@@ -1,21 +0,0 @@
-from facefusion.common_helper import is_macos, is_windows
-
-if is_windows():
-	import ctypes
-else:
-	import resource
-
-
-def limit_system_memory(system_memory_limit : int = 1) -> bool:
-	if is_macos():
-		system_memory_limit = system_memory_limit * (1024 ** 6)
-	else:
-		system_memory_limit = system_memory_limit * (1024 ** 3)
-	try:
-		if is_windows():
-			ctypes.windll.kernel32.SetProcessWorkingSetSize(-1, ctypes.c_size_t(system_memory_limit), ctypes.c_size_t(system_memory_limit)) #type:ignore[attr-defined]
-		else:
-			resource.setrlimit(resource.RLIMIT_DATA, (system_memory_limit, system_memory_limit))
-		return True
-	except Exception:
-		return False

facefusion/metadata.py

@@ -4,7 +4,7 @@ METADATA =\
 {
 	'name': 'FaceFusion',
 	'description': 'Industry leading face manipulation platform',
-	'version': '3.2.0',
+	'version': 'NEXT',
 	'license': 'OpenRAIL-AS',
 	'author': 'Henry Ruhs',
 	'url': 'https://facefusion.io'
@@ -12,6 +12,4 @@ METADATA =\
 
 
 def get(key : str) -> Optional[str]:
-	if key in METADATA:
-		return METADATA.get(key)
-	return None
+	return METADATA.get(key)

facefusion/model_helper.py

@@ -5,7 +5,7 @@ import onnx
 from facefusion.types import ModelInitializer
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def get_static_model_initializer(model_path : str) -> ModelInitializer:
 	model = onnx.load(model_path)
 	return onnx.numpy_helper.to_array(model.graph.initializer[-1])

facefusion/normalizer.py

@@ -1,17 +1,29 @@
 from typing import List, Optional
 
-from facefusion.types import Fps, Padding
+from facefusion.types import Color, Fps, Padding
 
 
-def normalize_padding(padding : Optional[List[int]]) -> Optional[Padding]:
-	if padding and len(padding) == 1:
-		return tuple([ padding[0] ] * 4) #type:ignore[return-value]
-	if padding and len(padding) == 2:
-		return tuple([ padding[0], padding[1], padding[0], padding[1] ]) #type:ignore[return-value]
-	if padding and len(padding) == 3:
-		return tuple([ padding[0], padding[1], padding[2], padding[1] ]) #type:ignore[return-value]
-	if padding and len(padding) == 4:
-		return tuple(padding) #type:ignore[return-value]
+def normalize_color(channels : Optional[List[int]]) -> Optional[Color]:
+	if channels and len(channels) == 1:
+		return tuple([ channels[0], channels[0], channels[0], 255 ]) #type:ignore[return-value]
+	if channels and len(channels) == 2:
+		return tuple([ channels[0], channels[1], channels[0], 255 ]) #type:ignore[return-value]
+	if channels and len(channels) == 3:
+		return tuple([ channels[0], channels[1], channels[2], 255 ]) #type:ignore[return-value]
+	if channels and len(channels) == 4:
+		return tuple(channels) #type:ignore[return-value]
+	return None
+
+
+def normalize_space(spaces : Optional[List[int]]) -> Optional[Padding]:
+	if spaces and len(spaces) == 1:
+		return tuple([spaces[0]] * 4) #type:ignore[return-value]
+	if spaces and len(spaces) == 2:
+		return tuple([ spaces[0], spaces[1], spaces[0], spaces[1] ]) #type:ignore[return-value]
+	if spaces and len(spaces) == 3:
+		return tuple([ spaces[0], spaces[1], spaces[2], spaces[1] ]) #type:ignore[return-value]
+	if spaces and len(spaces) == 4:
+		return tuple(spaces) #type:ignore[return-value]
 	return None
 
 

facefusion/process_manager.py

@@ -1,6 +1,4 @@
-from typing import Generator, List
-
-from facefusion.types import ProcessState, QueuePayload
+from facefusion.types import ProcessState
 
 PROCESS_STATE : ProcessState = 'pending'
 
@@ -45,9 +43,3 @@ def stop() -> None:
 
 def end() -> None:
 	set_process_state('pending')
-
-
-def manage(queue_payloads : List[QueuePayload]) -> Generator[QueuePayload, None, None]:
-	for query_payload in queue_payloads:
-		if is_processing():
-			yield query_payload

facefusion/processors/core.py

@@ -1,15 +1,10 @@
 import importlib
-import os
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from queue import Queue
 from types import ModuleType
 from typing import Any, List
 
-from tqdm import tqdm
-
-from facefusion import logger, state_manager, wording
+from facefusion import logger, translator
 from facefusion.exit_helper import hard_exit
-from facefusion.types import ProcessFrames, QueuePayload
+
 
 PROCESSORS_METHODS =\
 [
@@ -17,29 +12,26 @@ PROCESSORS_METHODS =\
 	'clear_inference_pool',
 	'register_args',
 	'apply_args',
+	'get_common_modules',
 	'pre_check',
 	'pre_process',
 	'post_process',
-	'get_reference_frame',
-	'process_frame',
-	'process_frames',
-	'process_image',
-	'process_video'
+	'process_frame'
 ]
 
 
 def load_processor_module(processor : str) -> Any:
 	try:
-		processor_module = importlib.import_module('facefusion.processors.modules.' + processor)
+		processor_module = importlib.import_module('facefusion.processors.modules.' + processor + '.core')
 		for method_name in PROCESSORS_METHODS:
 			if not hasattr(processor_module, method_name):
 				raise NotImplementedError
 	except ModuleNotFoundError as exception:
-		logger.error(wording.get('processor_not_loaded').format(processor = processor), __name__)
+		logger.error(translator.get('processor_not_loaded').format(processor = processor), __name__)
 		logger.debug(exception.msg, __name__)
 		hard_exit(1)
 	except NotImplementedError:
-		logger.error(wording.get('processor_not_implemented').format(processor = processor), __name__)
+		logger.error(translator.get('processor_not_implemented').format(processor = processor), __name__)
 		hard_exit(1)
 	return processor_module
 
@@ -51,49 +43,3 @@ def get_processors_modules(processors : List[str]) -> List[ModuleType]:
 		processor_module = load_processor_module(processor)
 		processor_modules.append(processor_module)
 	return processor_modules
-
-
-def multi_process_frames(source_paths : List[str], temp_frame_paths : List[str], process_frames : ProcessFrames) -> None:
-	queue_payloads = create_queue_payloads(temp_frame_paths)
-	with tqdm(total = len(queue_payloads), desc = wording.get('processing'), unit = 'frame', ascii = ' =', disable = state_manager.get_item('log_level') in [ 'warn', 'error' ]) as progress:
-		progress.set_postfix(execution_providers = state_manager.get_item('execution_providers'))
-		with ThreadPoolExecutor(max_workers = state_manager.get_item('execution_thread_count')) as executor:
-			futures = []
-			queue : Queue[QueuePayload] = create_queue(queue_payloads)
-			queue_per_future = max(len(queue_payloads) // state_manager.get_item('execution_thread_count') * state_manager.get_item('execution_queue_count'), 1)
-
-			while not queue.empty():
-				future = executor.submit(process_frames, source_paths, pick_queue(queue, queue_per_future), progress.update)
-				futures.append(future)
-
-			for future_done in as_completed(futures):
-				future_done.result()
-
-
-def create_queue(queue_payloads : List[QueuePayload]) -> Queue[QueuePayload]:
-	queue : Queue[QueuePayload] = Queue()
-	for queue_payload in queue_payloads:
-		queue.put(queue_payload)
-	return queue
-
-
-def pick_queue(queue : Queue[QueuePayload], queue_per_future : int) -> List[QueuePayload]:
-	queues = []
-	for _ in range(queue_per_future):
-		if not queue.empty():
-			queues.append(queue.get())
-	return queues
-
-
-def create_queue_payloads(temp_frame_paths : List[str]) -> List[QueuePayload]:
-	queue_payloads = []
-	temp_frame_paths = sorted(temp_frame_paths, key = os.path.basename)
-
-	for frame_number, frame_path in enumerate(temp_frame_paths):
-		frame_payload : QueuePayload =\
-		{
-			'frame_number': frame_number,
-			'frame_path': frame_path
-		}
-		queue_payloads.append(frame_payload)
-	return queue_payloads

facefusion/processors/live_portrait.py

@@ -63,15 +63,15 @@ def limit_expression(expression : LivePortraitExpression) -> LivePortraitExpress
 	return numpy.clip(expression, EXPRESSION_MIN, EXPRESSION_MAX)
 
 
-def limit_euler_angles(target_pitch : LivePortraitPitch, target_yaw : LivePortraitYaw, target_roll : LivePortraitRoll, output_pitch : LivePortraitPitch, output_yaw : LivePortraitYaw, output_roll : LivePortraitRoll) -> Tuple[LivePortraitPitch, LivePortraitYaw, LivePortraitRoll]:
-	pitch_min, pitch_max, yaw_min, yaw_max, roll_min, roll_max = calc_euler_limits(target_pitch, target_yaw, target_roll)
+def limit_angle(target_pitch : LivePortraitPitch, target_yaw : LivePortraitYaw, target_roll : LivePortraitRoll, output_pitch : LivePortraitPitch, output_yaw : LivePortraitYaw, output_roll : LivePortraitRoll) -> Tuple[LivePortraitPitch, LivePortraitYaw, LivePortraitRoll]:
+	pitch_min, pitch_max, yaw_min, yaw_max, roll_min, roll_max = calculate_euler_limits(target_pitch, target_yaw, target_roll)
 	output_pitch = numpy.clip(output_pitch, pitch_min, pitch_max)
 	output_yaw = numpy.clip(output_yaw, yaw_min, yaw_max)
 	output_roll = numpy.clip(output_roll, roll_min, roll_max)
 	return output_pitch, output_yaw, output_roll
 
 
-def calc_euler_limits(pitch : LivePortraitPitch, yaw : LivePortraitYaw, roll : LivePortraitRoll) -> Tuple[float, float, float, float, float, float]:
+def calculate_euler_limits(pitch : LivePortraitPitch, yaw : LivePortraitYaw, roll : LivePortraitRoll) -> Tuple[float, float, float, float, float, float]:
 	pitch_min = -30.0
 	pitch_max = 30.0
 	yaw_min = -60.0

facefusion/processors/modules/age_modifier.py

@@ -1,253 +0,0 @@
-from argparse import ArgumentParser
-from functools import lru_cache
-from typing import List
-
-import cv2
-import numpy
-
-import facefusion.choices
-import facefusion.jobs.job_manager
-import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import create_int_metavar
-from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.execution import has_execution_provider
-from facefusion.face_analyser import get_many_faces, get_one_face
-from facefusion.face_helper import merge_matrix, paste_back, scale_face_landmark_5, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
-from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.types import AgeModifierDirection, AgeModifierInputs
-from facefusion.program_helper import find_argument_group
-from facefusion.thread_helper import thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import match_frame_color, read_image, read_static_image, write_image
-
-
-@lru_cache(maxsize = None)
-def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
-	return\
-	{
-		'styleganex_age':
-		{
-			'hashes':
-			{
-				'age_modifier':
-				{
-					'url': resolve_download_url('models-3.1.0', 'styleganex_age.hash'),
-					'path': resolve_relative_path('../.assets/models/styleganex_age.hash')
-				}
-			},
-			'sources':
-			{
-				'age_modifier':
-				{
-					'url': resolve_download_url('models-3.1.0', 'styleganex_age.onnx'),
-					'path': resolve_relative_path('../.assets/models/styleganex_age.onnx')
-				}
-			},
-			'templates':
-			{
-				'target': 'ffhq_512',
-				'target_with_background': 'styleganex_384'
-			},
-			'sizes':
-			{
-				'target': (256, 256),
-				'target_with_background': (384, 384)
-			}
-		}
-	}
-
-
-def get_inference_pool() -> InferencePool:
-	model_names = [ state_manager.get_item('age_modifier_model') ]
-	model_source_set = get_model_options().get('sources')
-
-	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
-
-
-def clear_inference_pool() -> None:
-	model_names = [ state_manager.get_item('age_modifier_model') ]
-	inference_manager.clear_inference_pool(__name__, model_names)
-
-
-def get_model_options() -> ModelOptions:
-	model_name = state_manager.get_item('age_modifier_model')
-	return create_static_model_set('full').get(model_name)
-
-
-def register_args(program : ArgumentParser) -> None:
-	group_processors = find_argument_group(program, 'processors')
-	if group_processors:
-		group_processors.add_argument('--age-modifier-model', help = wording.get('help.age_modifier_model'), default = config.get_str_value('processors', 'age_modifier_model', 'styleganex_age'), choices = processors_choices.age_modifier_models)
-		group_processors.add_argument('--age-modifier-direction', help = wording.get('help.age_modifier_direction'), type = int, default = config.get_int_value('processors', 'age_modifier_direction', '0'), choices = processors_choices.age_modifier_direction_range, metavar = create_int_metavar(processors_choices.age_modifier_direction_range))
-		facefusion.jobs.job_store.register_step_keys([ 'age_modifier_model', 'age_modifier_direction' ])
-
-
-def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
-	apply_state_item('age_modifier_model', args.get('age_modifier_model'))
-	apply_state_item('age_modifier_direction', args.get('age_modifier_direction'))
-
-
-def pre_check() -> bool:
-	model_hash_set = get_model_options().get('hashes')
-	model_source_set = get_model_options().get('sources')
-
-	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
-
-
-def pre_process(mode : ProcessMode) -> bool:
-	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
-		return False
-	return True
-
-
-def post_process() -> None:
-	read_static_image.cache_clear()
-	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
-		clear_inference_pool()
-	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
-
-
-def modify_age(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	model_templates = get_model_options().get('templates')
-	model_sizes = get_model_options().get('sizes')
-	face_landmark_5 = target_face.landmark_set.get('5/68').copy()
-	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5, model_templates.get('target'), model_sizes.get('target'))
-	extend_face_landmark_5 = scale_face_landmark_5(face_landmark_5, 0.875)
-	extend_vision_frame, extend_affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, extend_face_landmark_5, model_templates.get('target_with_background'), model_sizes.get('target_with_background'))
-	extend_vision_frame_raw = extend_vision_frame.copy()
-	box_mask = create_static_box_mask(model_sizes.get('target_with_background'), state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
-	crop_masks =\
-	[
-		box_mask
-	]
-
-	if 'occlusion' in state_manager.get_item('face_mask_types'):
-		occlusion_mask = create_occlusion_mask(crop_vision_frame)
-		combined_matrix = merge_matrix([ extend_affine_matrix, cv2.invertAffineTransform(affine_matrix) ])
-		occlusion_mask = cv2.warpAffine(occlusion_mask, combined_matrix, model_sizes.get('target_with_background'))
-		crop_masks.append(occlusion_mask)
-
-	crop_vision_frame = prepare_vision_frame(crop_vision_frame)
-	extend_vision_frame = prepare_vision_frame(extend_vision_frame)
-	age_modifier_direction = numpy.array(numpy.interp(state_manager.get_item('age_modifier_direction'), [ -100, 100 ], [ 2.5, -2.5 ])).astype(numpy.float32)
-	extend_vision_frame = forward(crop_vision_frame, extend_vision_frame, age_modifier_direction)
-	extend_vision_frame = normalize_extend_frame(extend_vision_frame)
-	extend_vision_frame = match_frame_color(extend_vision_frame_raw, extend_vision_frame)
-	extend_affine_matrix *= (model_sizes.get('target')[0] * 4) / model_sizes.get('target_with_background')[0]
-	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
-	crop_mask = cv2.resize(crop_mask, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4))
-	paste_vision_frame = paste_back(temp_vision_frame, extend_vision_frame, crop_mask, extend_affine_matrix)
-	return paste_vision_frame
-
-
-def forward(crop_vision_frame : VisionFrame, extend_vision_frame : VisionFrame, age_modifier_direction : AgeModifierDirection) -> VisionFrame:
-	age_modifier = get_inference_pool().get('age_modifier')
-	age_modifier_inputs = {}
-
-	if has_execution_provider('coreml'):
-		age_modifier.set_providers([ facefusion.choices.execution_provider_set.get('cpu') ])
-
-	for age_modifier_input in age_modifier.get_inputs():
-		if age_modifier_input.name == 'target':
-			age_modifier_inputs[age_modifier_input.name] = crop_vision_frame
-		if age_modifier_input.name == 'target_with_background':
-			age_modifier_inputs[age_modifier_input.name] = extend_vision_frame
-		if age_modifier_input.name == 'direction':
-			age_modifier_inputs[age_modifier_input.name] = age_modifier_direction
-
-	with thread_semaphore():
-		crop_vision_frame = age_modifier.run(None, age_modifier_inputs)[0][0]
-
-	return crop_vision_frame
-
-
-def prepare_vision_frame(vision_frame : VisionFrame) -> VisionFrame:
-	vision_frame = vision_frame[:, :, ::-1] / 255.0
-	vision_frame = (vision_frame - 0.5) / 0.5
-	vision_frame = numpy.expand_dims(vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
-	return vision_frame
-
-
-def normalize_extend_frame(extend_vision_frame : VisionFrame) -> VisionFrame:
-	model_sizes = get_model_options().get('sizes')
-	extend_vision_frame = numpy.clip(extend_vision_frame, -1, 1)
-	extend_vision_frame = (extend_vision_frame + 1) / 2
-	extend_vision_frame = extend_vision_frame.transpose(1, 2, 0).clip(0, 255)
-	extend_vision_frame = (extend_vision_frame * 255.0)
-	extend_vision_frame = extend_vision_frame.astype(numpy.uint8)[:, :, ::-1]
-	extend_vision_frame = cv2.resize(extend_vision_frame, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4), interpolation = cv2.INTER_AREA)
-	return extend_vision_frame
-
-
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	return modify_age(target_face, temp_vision_frame)
-
-
-def process_frame(inputs : AgeModifierInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
-
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = modify_age(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = modify_age(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = modify_age(similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_path : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_path : str, target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(None, temp_frame_paths, process_frames)

facefusion/processors/modules/age_modifier/choices.py

@@ -0,0 +1,8 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_int_range
+from facefusion.processors.modules.age_modifier.types import AgeModifierModel
+
+age_modifier_models : List[AgeModifierModel] = list(get_args(AgeModifierModel))
+
+age_modifier_direction_range : Sequence[int] = create_int_range(-100, 100, 1)

facefusion/processors/modules/age_modifier/core.py

@@ -0,0 +1,300 @@
+from argparse import ArgumentParser
+from functools import lru_cache
+from types import ModuleType
+from typing import List
+
+import cv2
+import numpy
+
+import facefusion.choices
+import facefusion.jobs.job_manager
+import facefusion.jobs.job_store
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_int_metavar, get_middle
+from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
+from facefusion.face_creator import scale_face
+from facefusion.face_helper import merge_matrix, paste_back, scale_face_landmark_5, warp_face_by_face_landmark_5
+from facefusion.face_masker import create_box_mask, create_occlusion_mask
+from facefusion.face_selector import select_faces
+from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
+from facefusion.processors.modules.age_modifier import choices as age_modifier_choices
+from facefusion.processors.modules.age_modifier.types import AgeModifierDirection, AgeModifierInputs
+from facefusion.processors.types import ProcessorOutputs
+from facefusion.program_helper import find_argument_group
+from facefusion.thread_helper import thread_semaphore
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import match_frame_color, read_static_image, read_static_video_frame
+
+
+@lru_cache()
+def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
+	return\
+	{
+		'fran':
+		{
+			'__metadata__':
+			{
+				'vendor': 'ry-lu',
+				'license': 'mit',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.6.0', 'fran.hash'),
+					'path': resolve_relative_path('../.assets/models/fran.hash')
+				}
+			},
+			'sources':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.6.0', 'fran.onnx'),
+					'path': resolve_relative_path('../.assets/models/fran.onnx')
+				}
+			},
+			'templates':
+			{
+				'target': 'ffhq_512',
+			},
+			'sizes':
+			{
+				'target': (1024, 1024),
+			},
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'styleganex_age':
+		{
+			'__metadata__':
+			{
+				'vendor': 'williamyang1991',
+				'license': 'S-Lab-1.0',
+				'year': 2023
+			},
+			'hashes':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.1.0', 'styleganex_age.hash'),
+					'path': resolve_relative_path('../.assets/models/styleganex_age.hash')
+				}
+			},
+			'sources':
+			{
+				'age_modifier':
+				{
+					'url': resolve_download_url('models-3.1.0', 'styleganex_age.onnx'),
+					'path': resolve_relative_path('../.assets/models/styleganex_age.onnx')
+				}
+			},
+			'templates':
+			{
+				'target': 'ffhq_512',
+				'target_with_background': 'styleganex_384'
+			},
+			'sizes':
+			{
+				'target': (256, 256),
+				'target_with_background': (384, 384)
+			},
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
+		}
+	}
+
+
+def get_inference_pool() -> InferencePool:
+	model_names = [ state_manager.get_item('age_modifier_model') ]
+	model_source_set = get_model_options().get('sources')
+
+	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
+
+
+def clear_inference_pool() -> None:
+	model_names = [ state_manager.get_item('age_modifier_model') ]
+	inference_manager.clear_inference_pool(__name__, model_names)
+
+
+def get_model_options() -> ModelOptions:
+	model_name = state_manager.get_item('age_modifier_model')
+	return create_static_model_set('full').get(model_name)
+
+
+def register_args(program : ArgumentParser) -> None:
+	group_processors = find_argument_group(program, 'processors')
+	if group_processors:
+		group_processors.add_argument('--age-modifier-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'age_modifier_model', 'fran'), choices = age_modifier_choices.age_modifier_models)
+		group_processors.add_argument('--age-modifier-direction', help = translator.get('help.direction', __package__), type = int, default = config.get_int_value('processors', 'age_modifier_direction', '0'), choices = age_modifier_choices.age_modifier_direction_range, metavar = create_int_metavar(age_modifier_choices.age_modifier_direction_range))
+		facefusion.jobs.job_store.register_step_keys([ 'age_modifier_model', 'age_modifier_direction' ])
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('age_modifier_model', args.get('age_modifier_model'))
+	apply_state_item('age_modifier_direction', args.get('age_modifier_direction'))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
+def pre_check() -> bool:
+	model_hash_set = get_model_options().get('hashes')
+	model_source_set = get_model_options().get('sources')
+
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
+	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
+
+
+def pre_process(mode : ProcessMode) -> bool:
+	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
+		return False
+	return True
+
+
+def post_process() -> None:
+	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
+	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
+		clear_inference_pool()
+
+	if state_manager.get_item('video_memory_strategy') == 'strict':
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
+
+
+def modify_age(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	model_templates = get_model_options().get('templates')
+	model_sizes = get_model_options().get('sizes')
+	face_landmark_5 = target_face.landmark_set.get('5/68').copy()
+	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5, model_templates.get('target'), model_sizes.get('target'))
+
+	if state_manager.get_item('age_modifier_model') == 'fran':
+		box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+		crop_masks =\
+		[
+			box_mask
+		]
+
+		if 'occlusion' in state_manager.get_item('face_mask_types'):
+			occlusion_mask = create_occlusion_mask(crop_vision_frame)
+			crop_masks.append(occlusion_mask)
+
+		crop_vision_frame = prepare_vision_frame(crop_vision_frame)
+		target_age = numpy.mean(target_face.age)
+		age_modifier_direction = numpy.array([ target_age, target_age + state_manager.get_item('age_modifier_direction') ], dtype = numpy.float32) / 100
+		age_modifier_direction = age_modifier_direction.clip(0, 1)
+		crop_vision_frame = forward(crop_vision_frame, crop_vision_frame, age_modifier_direction)
+		crop_vision_frame = normalize_vision_frame(crop_vision_frame)
+		crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+		paste_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, crop_mask, affine_matrix)
+		return paste_vision_frame
+
+	if state_manager.get_item('age_modifier_model') == 'styleganex_age':
+		extend_face_landmark_5 = scale_face_landmark_5(face_landmark_5, 0.875)
+		extend_vision_frame, extend_affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, extend_face_landmark_5, model_templates.get('target_with_background'), model_sizes.get('target_with_background'))
+		extend_vision_frame_raw = extend_vision_frame.copy()
+		box_mask = create_box_mask(extend_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+		crop_masks =\
+		[
+			box_mask
+		]
+
+		if 'occlusion' in state_manager.get_item('face_mask_types'):
+			occlusion_mask = create_occlusion_mask(crop_vision_frame)
+			temp_matrix = merge_matrix([ extend_affine_matrix, cv2.invertAffineTransform(affine_matrix) ])
+			occlusion_mask = cv2.warpAffine(occlusion_mask, temp_matrix, model_sizes.get('target_with_background'))
+			crop_masks.append(occlusion_mask)
+
+		crop_vision_frame = prepare_vision_frame(crop_vision_frame)
+		extend_vision_frame = prepare_vision_frame(extend_vision_frame)
+		age_modifier_direction = numpy.array(numpy.interp(state_manager.get_item('age_modifier_direction'), [ -100, 100 ], [ 2.5, -2.5 ])).astype(numpy.float32)
+		extend_vision_frame = forward(crop_vision_frame, extend_vision_frame, age_modifier_direction)
+		extend_vision_frame = normalize_extend_frame(extend_vision_frame)
+		extend_vision_frame = match_frame_color(extend_vision_frame_raw, extend_vision_frame)
+		extend_affine_matrix *= (model_sizes.get('target')[0] * 4) / model_sizes.get('target_with_background')[0]
+		crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+		crop_mask = cv2.resize(crop_mask, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4))
+		paste_vision_frame = paste_back(temp_vision_frame, extend_vision_frame, crop_mask, extend_affine_matrix)
+		return paste_vision_frame
+
+	return temp_vision_frame
+
+
+def forward(crop_vision_frame : VisionFrame, extend_vision_frame : VisionFrame, age_modifier_direction : AgeModifierDirection) -> VisionFrame:
+	age_modifier = get_inference_pool().get('age_modifier')
+	age_modifier_inputs = {}
+
+	for age_modifier_input in age_modifier.get_inputs():
+		if age_modifier_input.name == 'target':
+			age_modifier_inputs[age_modifier_input.name] = crop_vision_frame
+		if age_modifier_input.name == 'target_with_background':
+			age_modifier_inputs[age_modifier_input.name] = extend_vision_frame
+		if age_modifier_input.name == 'direction':
+			age_modifier_inputs[age_modifier_input.name] = age_modifier_direction
+
+	with thread_semaphore():
+		crop_vision_frame = age_modifier.run(None, age_modifier_inputs)[0][0]
+
+	return crop_vision_frame
+
+
+def prepare_vision_frame(vision_frame : VisionFrame) -> VisionFrame:
+	model_mean = get_model_options().get('mean')
+	model_standard_deviation = get_model_options().get('standard_deviation')
+	vision_frame = vision_frame[:, :, ::-1] / 255.0
+	vision_frame = (vision_frame - model_mean) / model_standard_deviation
+	vision_frame = numpy.expand_dims(vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
+	return vision_frame
+
+
+def normalize_vision_frame(vision_frame : VisionFrame) -> VisionFrame:
+	model_mean = get_model_options().get('mean')
+	model_standard_deviation = get_model_options().get('standard_deviation')
+	vision_frame = vision_frame.transpose(1, 2, 0)
+	vision_frame = vision_frame * model_standard_deviation + model_mean
+	vision_frame = vision_frame.clip(0, 1)
+	vision_frame = vision_frame[:, :, ::-1] * 255
+	return vision_frame
+
+
+def normalize_extend_frame(extend_vision_frame : VisionFrame) -> VisionFrame:
+	model_sizes = get_model_options().get('sizes')
+	extend_vision_frame = numpy.clip(extend_vision_frame, -1, 1)
+	extend_vision_frame = (extend_vision_frame + 1) / 2
+	extend_vision_frame = extend_vision_frame.transpose(1, 2, 0).clip(0, 255)
+	extend_vision_frame = (extend_vision_frame * 255.0)
+	extend_vision_frame = extend_vision_frame.astype(numpy.uint8)[:, :, ::-1]
+	extend_vision_frame = cv2.resize(extend_vision_frame, (model_sizes.get('target')[0] * 4, model_sizes.get('target')[1] * 4), interpolation = cv2.INTER_AREA)
+	return extend_vision_frame
+
+
+def process_frame(inputs : AgeModifierInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
+
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
+
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = modify_age(target_face, temp_vision_frame)
+
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/age_modifier/locales.py

@@ -0,0 +1,18 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for aging the face',
+			'direction': 'specify the direction in which the age should be modified'
+		},
+		'uis':
+		{
+			'direction_slider': 'AGE MODIFIER DIRECTION',
+			'model_dropdown': 'AGE MODIFIER MODEL'
+		}
+	}
+}

facefusion/processors/modules/age_modifier/types.py

@@ -0,0 +1,18 @@
+from typing import Any, List, Literal, TypeAlias, TypedDict
+
+from numpy.typing import NDArray
+
+from facefusion.types import Mask, VisionFrame
+
+AgeModifierInputs = TypedDict('AgeModifierInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+AgeModifierModel = Literal['fran', 'styleganex_age']
+
+AgeModifierDirection : TypeAlias = NDArray[Any]

facefusion/processors/modules/background_remover/choices.py

@@ -0,0 +1,8 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_int_range
+from facefusion.processors.modules.background_remover.types import BackgroundRemoverModel
+
+background_remover_models : List[BackgroundRemoverModel] = list(get_args(BackgroundRemoverModel))
+
+background_remover_color_range : Sequence[int] = create_int_range(0, 255, 1)

facefusion/processors/modules/background_remover/core.py

@@ -0,0 +1,661 @@
+from argparse import ArgumentParser
+from functools import lru_cache, partial
+from types import ModuleType
+from typing import List, Tuple
+
+import cv2
+import numpy
+
+import facefusion.choices
+import facefusion.jobs.job_manager
+import facefusion.jobs.job_store
+from facefusion import config, content_analyser, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import is_macos, is_windows
+from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
+from facefusion.execution import has_execution_provider
+from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
+from facefusion.normalizer import normalize_color
+from facefusion.processors.modules.background_remover import choices as background_remover_choices
+from facefusion.processors.modules.background_remover.types import BackgroundRemoverInputs
+from facefusion.processors.types import ProcessorOutputs
+from facefusion.program_helper import find_argument_group
+from facefusion.sanitizer import sanitize_int_range
+from facefusion.thread_helper import thread_semaphore
+from facefusion.types import ApplyStateItem, Args, DownloadScope, InferencePool, InferenceProvider, Mask, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_video_frame
+
+
+@lru_cache()
+def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
+	return\
+	{
+		'ben_2':
+		{
+			'__metadata__':
+			{
+				'vendor': 'PramaLLC',
+				'license': 'MIT',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'ben_2.hash'),
+					'path': resolve_relative_path('../.assets/models/ben_2.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'ben_2.onnx'),
+					'path': resolve_relative_path('../.assets/models/ben_2.onnx')
+				}
+			},
+			'type': 'ben',
+			'size': (1024, 1024),
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'birefnet_general':
+		{
+			'__metadata__':
+			{
+				'vendor': 'ZhengPeng7',
+				'license': 'MIT',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'birefnet_general.hash'),
+					'path': resolve_relative_path('../.assets/models/birefnet_general.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'birefnet_general.onnx'),
+					'path': resolve_relative_path('../.assets/models/birefnet_general.onnx')
+				}
+			},
+			'type': 'birefnet',
+			'size': (1024, 1024),
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'birefnet_portrait':
+		{
+			'__metadata__':
+			{
+				'vendor': 'ZhengPeng7',
+				'license': 'MIT',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'birefnet_portrait.hash'),
+					'path': resolve_relative_path('../.assets/models/birefnet_portrait.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'birefnet_portrait.onnx'),
+					'path': resolve_relative_path('../.assets/models/birefnet_portrait.onnx')
+				}
+			},
+			'type': 'birefnet',
+			'size': (1024, 1024),
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'corridor_key_1024':
+		{
+			'__metadata__':
+			{
+				'vendor': 'nikopueringer',
+				'license': 'Non-Commercial',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_1024.hash'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_1024.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_1024.onnx'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_1024.onnx')
+				}
+			},
+			'type': 'corridor_key',
+			'size': (1024, 1024),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'corridor_key_2048':
+		{
+			'__metadata__':
+			{
+				'vendor': 'nikopueringer',
+				'license': 'Non-Commercial',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_2048.hash'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_2048.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.6.0', 'corridor_key_2048.onnx'),
+					'path': resolve_relative_path('../.assets/models/corridor_key_2048.onnx')
+				}
+			},
+			'type': 'corridor_key',
+			'size': (2048, 2048),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'isnet_general':
+		{
+			'__metadata__':
+			{
+				'vendor': 'xuebinqin',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'isnet_general.hash'),
+					'path': resolve_relative_path('../.assets/models/isnet_general.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'isnet_general.onnx'),
+					'path': resolve_relative_path('../.assets/models/isnet_general.onnx')
+				}
+			},
+			'type': 'isnet',
+			'size': (1024, 1024),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'modnet':
+		{
+			'__metadata__':
+			{
+				'vendor': 'ZHKKKe',
+				'license': 'Apache-2.0',
+				'year': 2020
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'modnet.hash'),
+					'path': resolve_relative_path('../.assets/models/modnet.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'modnet.onnx'),
+					'path': resolve_relative_path('../.assets/models/modnet.onnx')
+				}
+			},
+			'type': 'modnet',
+			'size': (512, 512),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
+		},
+		'ormbg':
+		{
+			'__metadata__':
+			{
+				'vendor': 'schirrmacher',
+				'license': 'Apache-2.0',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'ormbg.hash'),
+					'path': resolve_relative_path('../.assets/models/ormbg.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'ormbg.onnx'),
+					'path': resolve_relative_path('../.assets/models/ormbg.onnx')
+				}
+			},
+			'type': 'ormbg',
+			'size': (1024, 1024),
+			'mean': [ 0.0, 0.0, 0.0 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'rmbg_1.4':
+		{
+			'__metadata__':
+			{
+				'vendor': 'Bria',
+				'license': 'Non-Commercial',
+				'year': 2023
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'rmbg_1.4.hash'),
+					'path': resolve_relative_path('../.assets/models/rmbg_1.4.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'rmbg_1.4.onnx'),
+					'path': resolve_relative_path('../.assets/models/rmbg_1.4.onnx')
+				}
+			},
+			'type': 'rmbg',
+			'size': (1024, 1024),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 1.0, 1.0, 1.0 ]
+		},
+		'rmbg_2.0':
+		{
+			'__metadata__':
+			{
+				'vendor': 'Bria',
+				'license': 'Non-Commercial',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'rmbg_2.0.hash'),
+					'path': resolve_relative_path('../.assets/models/rmbg_2.0.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'rmbg_2.0.onnx'),
+					'path': resolve_relative_path('../.assets/models/rmbg_2.0.onnx')
+				}
+			},
+			'type': 'rmbg',
+			'size': (1024, 1024),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'silueta':
+		{
+			'__metadata__':
+			{
+				'vendor': 'Kikedao',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'silueta.hash'),
+					'path': resolve_relative_path('../.assets/models/silueta.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'silueta.onnx'),
+					'path': resolve_relative_path('../.assets/models/silueta.onnx')
+				}
+			},
+			'type': 'silueta',
+			'size': (320, 320),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'u2net_cloth':
+		{
+			'__metadata__':
+			{
+				'vendor': 'levindabhi',
+				'license': 'MIT',
+				'year': 2021
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_cloth.hash'),
+					'path': resolve_relative_path('../.assets/models/u2net_cloth.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_cloth.onnx'),
+					'path': resolve_relative_path('../.assets/models/u2net_cloth.onnx')
+				}
+			},
+			'type': 'u2net_cloth',
+			'size': (768, 768),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'u2net_general':
+		{
+			'__metadata__':
+			{
+				'vendor': 'xuebinqin',
+				'license': 'Apache-2.0',
+				'year': 2020
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_general.hash'),
+					'path': resolve_relative_path('../.assets/models/u2net_general.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_general.onnx'),
+					'path': resolve_relative_path('../.assets/models/u2net_general.onnx')
+				}
+			},
+			'type': 'u2net',
+			'size': (320, 320),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'u2net_human':
+		{
+			'__metadata__':
+			{
+				'vendor': 'xuebinqin',
+				'license': 'Apache-2.0',
+				'year': 2021
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_human.hash'),
+					'path': resolve_relative_path('../.assets/models/u2net_human.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2net_human.onnx'),
+					'path': resolve_relative_path('../.assets/models/u2net_human.onnx')
+				}
+			},
+			'type': 'u2net',
+			'size': (320, 320),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		},
+		'u2netp':
+		{
+			'__metadata__':
+			{
+				'vendor': 'xuebinqin',
+				'license': 'Apache-2.0',
+				'year': 2021
+			},
+			'hashes':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2netp.hash'),
+					'path': resolve_relative_path('../.assets/models/u2netp.hash')
+				}
+			},
+			'sources':
+			{
+				'background_remover':
+				{
+					'url': resolve_download_url('models-3.5.0', 'u2netp.onnx'),
+					'path': resolve_relative_path('../.assets/models/u2netp.onnx')
+				}
+			},
+			'type': 'u2netp',
+			'size': (320, 320),
+			'mean': [ 0.485, 0.456, 0.406 ],
+			'standard_deviation': [ 0.229, 0.224, 0.225 ]
+		}
+	}
+
+
+def get_inference_pool() -> InferencePool:
+	model_names = [ state_manager.get_item('background_remover_model') ]
+	model_source_set = get_model_options().get('sources')
+
+	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
+
+
+def clear_inference_pool() -> None:
+	model_names = [ state_manager.get_item('background_remover_model') ]
+	inference_manager.clear_inference_pool(__name__, model_names)
+
+
+def override_inference_providers() -> List[InferenceProvider]:
+	model_type = get_model_options().get('type')
+
+	if is_macos() and has_execution_provider('coreml') or is_windows() and has_execution_provider('directml') and model_type == 'corridor_key':
+		return [ facefusion.choices.execution_provider_set.get('cpu') ]
+
+	return []
+
+
+def get_model_options() -> ModelOptions:
+	model_name = state_manager.get_item('background_remover_model')
+	return create_static_model_set('full').get(model_name)
+
+
+def register_args(program : ArgumentParser) -> None:
+	group_processors = find_argument_group(program, 'processors')
+	if group_processors:
+		group_processors.add_argument('--background-remover-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'background_remover_model', 'modnet'), choices = background_remover_choices.background_remover_models)
+		group_processors.add_argument('--background-remover-fill-color', help = translator.get('help.fill_color', __package__), type = partial(sanitize_int_range, int_range = background_remover_choices.background_remover_color_range), default = config.get_int_list('processors', 'background_remover_fill_color', '0 0 0 0'), nargs = '+')
+		group_processors.add_argument('--background-remover-despill-color', help = translator.get('help.despill_color', __package__), type = partial(sanitize_int_range, int_range = background_remover_choices.background_remover_color_range), default = config.get_int_list('processors', 'background_remover_despill_color', '0 0 0 0'), nargs = '+')
+		facefusion.jobs.job_store.register_step_keys([ 'background_remover_model', 'background_remover_fill_color', 'background_remover_despill_color' ])
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('background_remover_model', args.get('background_remover_model'))
+	apply_state_item('background_remover_fill_color', normalize_color(args.get('background_remover_fill_color')))
+	apply_state_item('background_remover_despill_color', normalize_color(args.get('background_remover_despill_color')))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser ]
+
+
+def pre_check() -> bool:
+	model_hash_set = get_model_options().get('hashes')
+	model_source_set = get_model_options().get('sources')
+
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
+	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
+
+
+def pre_process(mode : ProcessMode) -> bool:
+	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
+		return False
+	return True
+
+
+def post_process() -> None:
+	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
+	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
+		clear_inference_pool()
+
+	if state_manager.get_item('video_memory_strategy') == 'strict':
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
+
+
+def remove_background(temp_vision_frame : VisionFrame) -> Tuple[VisionFrame, Mask]:
+	model_type = get_model_options().get('type')
+
+	if model_type == 'corridor_key':
+		remove_vision_mask, remove_vision_frame = forward_corridor_key(prepare_temp_frame(temp_vision_frame))
+		remove_vision_frame = numpy.squeeze(remove_vision_frame).transpose(1, 2, 0)
+		remove_vision_frame = numpy.clip(remove_vision_frame * 255, 0, 255).astype(numpy.uint8)
+		temp_vision_frame = cv2.resize(remove_vision_frame[:, :, ::-1], temp_vision_frame.shape[:2][::-1])
+	else:
+		remove_vision_mask = forward(prepare_temp_frame(temp_vision_frame))
+
+	remove_vision_mask = normalize_vision_mask(remove_vision_mask)
+	remove_vision_mask = cv2.resize(remove_vision_mask, temp_vision_frame.shape[:2][::-1])
+	temp_vision_frame = apply_despill_color(temp_vision_frame)
+	temp_vision_frame = apply_fill_color(temp_vision_frame, remove_vision_mask)
+	return temp_vision_frame, remove_vision_mask
+
+
+def forward(temp_vision_frame : VisionFrame) -> VisionFrame:
+	background_remover = get_inference_pool().get('background_remover')
+	model_type = get_model_options().get('type')
+
+	with thread_semaphore():
+		remove_vision_frame = background_remover.run(None,
+		{
+			'input': temp_vision_frame
+		})[0]
+
+		if model_type == 'u2net_cloth':
+			remove_vision_frame = numpy.argmax(remove_vision_frame, axis = 1)
+
+	return remove_vision_frame
+
+
+def forward_corridor_key(temp_vision_frame : VisionFrame) -> Tuple[Mask, VisionFrame]:
+	background_remover = get_inference_pool().get('background_remover')
+
+	with thread_semaphore():
+		remove_vision_mask, remove_vision_frame = background_remover.run(None,
+		{
+			'input': temp_vision_frame
+		})
+
+	return remove_vision_mask, remove_vision_frame
+
+
+def prepare_temp_frame(temp_vision_frame : VisionFrame) -> VisionFrame:
+	model_type = get_model_options().get('type')
+	model_size = get_model_options().get('size')
+	model_mean = get_model_options().get('mean')
+	model_standard_deviation = get_model_options().get('standard_deviation')
+
+	if model_type == 'corridor_key':
+		coarse_color = temp_vision_frame[:, :, ::-1].astype(numpy.float32) / 255.0
+		coarse_bias = coarse_color[:, :, 1] - numpy.maximum(coarse_color[:, :, 0], coarse_color[:, :, 2])
+		coarse_vision_mask = cv2.resize(1.0 - numpy.clip(coarse_bias * 2.0, 0, 1), model_size)[:, :, numpy.newaxis]
+
+	temp_vision_frame = cv2.resize(temp_vision_frame, model_size)
+	temp_vision_frame = temp_vision_frame[:, :, ::-1] / 255.0
+	temp_vision_frame = (temp_vision_frame - model_mean) / model_standard_deviation
+
+	if model_type == 'corridor_key':
+		temp_vision_frame = numpy.concatenate([ temp_vision_frame, coarse_vision_mask ], axis = 2)
+
+	temp_vision_frame = temp_vision_frame.transpose(2, 0, 1)
+	temp_vision_frame = numpy.expand_dims(temp_vision_frame, axis = 0).astype(numpy.float32)
+	return temp_vision_frame
+
+
+def normalize_vision_mask(temp_vision_mask : Mask) -> Mask:
+	temp_vision_mask = numpy.squeeze(temp_vision_mask).clip(0, 1) * 255
+	temp_vision_mask = numpy.clip(temp_vision_mask, 0, 255).astype(numpy.uint8)
+	return temp_vision_mask
+
+
+def apply_fill_color(temp_vision_frame : VisionFrame, temp_vision_mask : Mask) -> VisionFrame:
+	background_remover_fill_color = state_manager.get_item('background_remover_fill_color')
+	temp_vision_mask = temp_vision_mask.astype(numpy.float32) / 255
+	temp_vision_mask = numpy.expand_dims(temp_vision_mask, axis = 2)
+	temp_vision_mask = (1 - temp_vision_mask) * background_remover_fill_color[-1] / 255
+	fill_vision_frame = numpy.zeros_like(temp_vision_frame)
+	fill_vision_frame[:, :, 0] = background_remover_fill_color[2]
+	fill_vision_frame[:, :, 1] = background_remover_fill_color[1]
+	fill_vision_frame[:, :, 2] = background_remover_fill_color[0]
+	temp_vision_frame = temp_vision_frame * (1 - temp_vision_mask) + fill_vision_frame * temp_vision_mask
+	temp_vision_frame = temp_vision_frame.astype(numpy.uint8)
+	return temp_vision_frame
+
+
+def apply_despill_color(temp_vision_frame : VisionFrame) -> VisionFrame:
+	background_remover_despill_color = state_manager.get_item('background_remover_despill_color')
+	temp_vision_frame = temp_vision_frame.astype(numpy.float32)
+	color_alpha = background_remover_despill_color[3] / 255.0
+	despill_vision_frame = numpy.zeros_like(temp_vision_frame)
+	despill_vision_frame[:, :, 0] = background_remover_despill_color[2]
+	despill_vision_frame[:, :, 1] = background_remover_despill_color[1]
+	despill_vision_frame[:, :, 2] = background_remover_despill_color[0]
+	color_weight = despill_vision_frame / numpy.maximum(numpy.max(background_remover_despill_color[:3]), 1)
+	color_limit = numpy.roll(temp_vision_frame, 1, 2) + numpy.roll(temp_vision_frame, -1, 2)
+	limit_vision_frame = numpy.minimum(temp_vision_frame, color_limit * 0.5)
+	temp_vision_frame = temp_vision_frame + (limit_vision_frame - temp_vision_frame) * color_alpha * color_weight
+	temp_vision_frame = temp_vision_frame.astype(numpy.uint8)
+	return temp_vision_frame
+
+
+def process_frame(inputs : BackgroundRemoverInputs) -> ProcessorOutputs:
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_frame, temp_vision_mask = remove_background(temp_vision_frame)
+	temp_vision_mask = numpy.minimum.reduce([ temp_vision_mask, inputs.get('temp_vision_mask') ])
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/background_remover/locales.py

@@ -0,0 +1,26 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for removing the background',
+			'fill_color': 'apply red, green, blue and alpha values to the background',
+			'despill_color': 'remove red, green, blue and alpha values from the foreground'
+		},
+		'uis':
+		{
+			'model_dropdown': 'BACKGROUND REMOVER MODEL',
+			'fill_color_red_number': 'FILL COLOR RED',
+			'fill_color_green_number': 'FILL COLOR GREEN',
+			'fill_color_blue_number': 'FILL COLOR BLUE',
+			'fill_color_alpha_number': 'FILL COLOR ALPHA',
+			'despill_color_red_number': 'DESPILL COLOR RED',
+			'despill_color_green_number': 'DESPILL COLOR GREEN',
+			'despill_color_blue_number': 'DESPILL COLOR BLUE',
+			'despill_color_alpha_number': 'DESPILL COLOR ALPHA'
+		}
+	}
+}

facefusion/processors/modules/background_remover/types.py

@@ -0,0 +1,12 @@
+from typing import List, Literal, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+BackgroundRemoverInputs = TypedDict('BackgroundRemoverInputs',
+{
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+BackgroundRemoverModel = Literal['ben_2', 'birefnet_general', 'birefnet_portrait', 'corridor_key_1024', 'corridor_key_2048', 'isnet_general', 'modnet', 'ormbg', 'rmbg_1.4', 'rmbg_2.0', 'silueta', 'u2net_cloth', 'u2net_general', 'u2net_human', 'u2netp']

facefusion/processors/modules/deep_swapper/choices.py

@@ -1,12 +1,12 @@
 from typing import List, Sequence
 
-from facefusion.common_helper import create_float_range, create_int_range
+from facefusion.common_helper import create_int_range
 from facefusion.filesystem import get_file_name, resolve_file_paths, resolve_relative_path
-from facefusion.processors.types import AgeModifierModel, DeepSwapperModel, ExpressionRestorerModel, FaceDebuggerItem, FaceEditorModel, FaceEnhancerModel, FaceSwapperModel, FaceSwapperSet, FrameColorizerModel, FrameEnhancerModel, LipSyncerModel
+from facefusion.processors.modules.deep_swapper.types import DeepSwapperModel
 
-age_modifier_models : List[AgeModifierModel] = [ 'styleganex_age' ]
 deep_swapper_models : List[DeepSwapperModel] =\
 [
+	'druuzil/adam_levine_320',
 	'druuzil/adrianne_palicki_384',
 	'druuzil/agnetha_falskog_224',
 	'druuzil/alan_ritchson_320',
@@ -14,6 +14,7 @@ deep_swapper_models : List[DeepSwapperModel] =\
 	'druuzil/amber_midthunder_320',
 	'druuzil/andras_arato_384',
 	'druuzil/andrew_tate_320',
+	'druuzil/angelina_jolie_384',
 	'druuzil/anne_hathaway_320',
 	'druuzil/anya_chalotra_320',
 	'druuzil/arnold_schwarzenegger_320',
@@ -21,6 +22,7 @@ deep_swapper_models : List[DeepSwapperModel] =\
 	'druuzil/benjamin_stiller_384',
 	'druuzil/bradley_pitt_224',
 	'druuzil/brie_larson_384',
+	'druuzil/bruce_campbell_384',
 	'druuzil/bryan_cranston_320',
 	'druuzil/catherine_blanchett_352',
 	'druuzil/christian_bale_320',
@@ -50,6 +52,7 @@ deep_swapper_models : List[DeepSwapperModel] =\
 	'druuzil/hugh_jackman_384',
 	'druuzil/idris_elba_320',
 	'druuzil/jack_nicholson_320',
+	'druuzil/james_carrey_384',
 	'druuzil/james_mcavoy_320',
 	'druuzil/james_varney_320',
 	'druuzil/jason_momoa_320',
@@ -61,6 +64,7 @@ deep_swapper_models : List[DeepSwapperModel] =\
 	'druuzil/kate_beckinsale_384',
 	'druuzil/laurence_fishburne_384',
 	'druuzil/lili_reinhart_320',
+	'druuzil/luke_evans_384',
 	'druuzil/mads_mikkelsen_384',
 	'druuzil/mary_winstead_320',
 	'druuzil/margaret_qualley_384',
@@ -69,13 +73,16 @@ deep_swapper_models : List[DeepSwapperModel] =\
 	'druuzil/michael_fox_320',
 	'druuzil/millie_bobby_brown_320',
 	'druuzil/morgan_freeman_320',
-	'druuzil/patrick_stewart_320',
+	'druuzil/patrick_stewart_224',
+	'druuzil/rachel_weisz_384',
 	'druuzil/rebecca_ferguson_320',
 	'druuzil/scarlett_johansson_320',
+	'druuzil/shannen_doherty_384',
 	'druuzil/seth_macfarlane_384',
 	'druuzil/thomas_cruise_320',
 	'druuzil/thomas_hanks_384',
 	'druuzil/william_murray_384',
+	'druuzil/zoe_saldana_384',
 	'edel/emma_roberts_224',
 	'edel/ivanka_trump_224',
 	'edel/lize_dzjabrailova_224',
@@ -166,47 +173,4 @@ if custom_model_file_paths:
 		model_id = '/'.join([ 'custom', get_file_name(model_file_path) ])
 		deep_swapper_models.append(model_id)
 
-expression_restorer_models : List[ExpressionRestorerModel] = [ 'live_portrait' ]
-face_debugger_items : List[FaceDebuggerItem] = [ 'bounding-box', 'face-landmark-5', 'face-landmark-5/68', 'face-landmark-68', 'face-landmark-68/5', 'face-mask', 'face-detector-score', 'face-landmarker-score', 'age', 'gender', 'race' ]
-face_editor_models : List[FaceEditorModel] = [ 'live_portrait' ]
-face_enhancer_models : List[FaceEnhancerModel] = [ 'codeformer', 'gfpgan_1.2', 'gfpgan_1.3', 'gfpgan_1.4', 'gpen_bfr_256', 'gpen_bfr_512', 'gpen_bfr_1024', 'gpen_bfr_2048', 'restoreformer_plus_plus' ]
-face_swapper_set : FaceSwapperSet =\
-{
-	'blendswap_256': [ '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
-	'ghost_1_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
-	'ghost_2_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
-	'ghost_3_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
-	'hififace_unofficial_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
-	'inswapper_128': [ '128x128', '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
-	'inswapper_128_fp16': [ '128x128', '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
-	'simswap_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
-	'simswap_unofficial_512': [ '512x512', '768x768', '1024x1024' ],
-	'uniface_256': [ '256x256', '512x512', '768x768', '1024x1024' ]
-}
-face_swapper_models : List[FaceSwapperModel] = list(face_swapper_set.keys())
-frame_colorizer_models : List[FrameColorizerModel] = [ 'ddcolor', 'ddcolor_artistic', 'deoldify', 'deoldify_artistic', 'deoldify_stable' ]
-frame_colorizer_sizes : List[str] = [ '192x192', '256x256', '384x384', '512x512' ]
-frame_enhancer_models : List[FrameEnhancerModel] = [ 'clear_reality_x4', 'lsdir_x4', 'nomos8k_sc_x4', 'real_esrgan_x2', 'real_esrgan_x2_fp16', 'real_esrgan_x4', 'real_esrgan_x4_fp16', 'real_esrgan_x8', 'real_esrgan_x8_fp16', 'real_hatgan_x4', 'real_web_photo_x4', 'realistic_rescaler_x4', 'remacri_x4', 'siax_x4', 'span_kendata_x4', 'swin2_sr_x4', 'ultra_sharp_x4' ]
-lip_syncer_models : List[LipSyncerModel] = [ 'wav2lip_96', 'wav2lip_gan_96' ]
-
-age_modifier_direction_range : Sequence[int] = create_int_range(-100, 100, 1)
 deep_swapper_morph_range : Sequence[int] = create_int_range(0, 100, 1)
-expression_restorer_factor_range : Sequence[int] = create_int_range(0, 100, 1)
-face_editor_eyebrow_direction_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_eye_gaze_horizontal_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_eye_gaze_vertical_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_eye_open_ratio_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_lip_open_ratio_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_grim_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_pout_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_purse_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_smile_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_position_horizontal_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_mouth_position_vertical_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_head_pitch_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_head_yaw_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_editor_head_roll_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
-face_enhancer_blend_range : Sequence[int] = create_int_range(0, 100, 1)
-face_enhancer_weight_range : Sequence[float] = create_float_range(0.0, 1.0, 0.05)
-frame_colorizer_blend_range : Sequence[int] = create_int_range(0, 100, 1)
-frame_enhancer_blend_range : Sequence[int] = create_int_range(0, 100, 1)

facefusion/processors/modules/deep_swapper/core.py

@@ -1,5 +1,6 @@
 from argparse import ArgumentParser
 from functools import lru_cache
+from types import ModuleType
 from typing import List, Tuple
 
 import cv2
@@ -8,31 +9,31 @@ from cv2.typing import Size
 
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import create_int_metavar
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_int_metavar, get_middle
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url_by_provider
-from facefusion.face_analyser import get_many_faces, get_one_face
+from facefusion.face_creator import scale_face
 from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_region_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
+from facefusion.face_masker import create_area_mask, create_box_mask, create_occlusion_mask, create_region_mask
+from facefusion.face_selector import select_faces
 from facefusion.filesystem import get_file_name, in_directory, is_image, is_video, resolve_file_paths, resolve_relative_path, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.types import DeepSwapperInputs, DeepSwapperMorph
+from facefusion.processors.modules.deep_swapper import choices as deep_swapper_choices
+from facefusion.processors.modules.deep_swapper.types import DeepSwapperInputs, DeepSwapperMorph
+from facefusion.processors.types import ProcessorOutputs
 from facefusion.program_helper import find_argument_group
 from facefusion.thread_helper import thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, Mask, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import conditional_match_frame_color, read_image, read_static_image, write_image
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, Mask, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import conditional_match_frame_color, read_static_image, read_static_video_frame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	model_config = []
 
 	if download_scope == 'full':
 		model_config.extend(
 		[
+			('druuzil', 'adam_levine_320'),
 			('druuzil', 'adrianne_palicki_384'),
 			('druuzil', 'agnetha_falskog_224'),
 			('druuzil', 'alan_ritchson_320'),
@@ -40,6 +41,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			('druuzil', 'amber_midthunder_320'),
 			('druuzil', 'andras_arato_384'),
 			('druuzil', 'andrew_tate_320'),
+			('druuzil', 'angelina_jolie_384'),
 			('druuzil', 'anne_hathaway_320'),
 			('druuzil', 'anya_chalotra_320'),
 			('druuzil', 'arnold_schwarzenegger_320'),
@@ -47,6 +49,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			('druuzil', 'benjamin_stiller_384'),
 			('druuzil', 'bradley_pitt_224'),
 			('druuzil', 'brie_larson_384'),
+			('druuzil', 'bruce_campbell_384'),
 			('druuzil', 'bryan_cranston_320'),
 			('druuzil', 'catherine_blanchett_352'),
 			('druuzil', 'christian_bale_320'),
@@ -76,6 +79,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			('druuzil', 'hugh_jackman_384'),
 			('druuzil', 'idris_elba_320'),
 			('druuzil', 'jack_nicholson_320'),
+			('druuzil', 'james_carrey_384'),
 			('druuzil', 'james_mcavoy_320'),
 			('druuzil', 'james_varney_320'),
 			('druuzil', 'jason_momoa_320'),
@@ -87,6 +91,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			('druuzil', 'kate_beckinsale_384'),
 			('druuzil', 'laurence_fishburne_384'),
 			('druuzil', 'lili_reinhart_320'),
+			('druuzil', 'luke_evans_384'),
 			('druuzil', 'mads_mikkelsen_384'),
 			('druuzil', 'mary_winstead_320'),
 			('druuzil', 'margaret_qualley_384'),
@@ -95,13 +100,16 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			('druuzil', 'michael_fox_320'),
 			('druuzil', 'millie_bobby_brown_320'),
 			('druuzil', 'morgan_freeman_320'),
-			('druuzil', 'patrick_stewart_320'),
+			('druuzil', 'patrick_stewart_224'),
+			('druuzil', 'rachel_weisz_384'),
 			('druuzil', 'rebecca_ferguson_320'),
 			('druuzil', 'scarlett_johansson_320'),
+			('druuzil', 'shannen_doherty_384'),
 			('druuzil', 'seth_macfarlane_384'),
 			('druuzil', 'thomas_cruise_320'),
 			('druuzil', 'thomas_hanks_384'),
 			('druuzil', 'william_murray_384'),
+			('druuzil', 'zoe_saldana_384'),
 			('edel', 'emma_roberts_224'),
 			('edel', 'ivanka_trump_224'),
 			('edel', 'lize_dzjabrailova_224'),
@@ -269,8 +277,8 @@ def get_model_size() -> Size:
 def register_args(program : ArgumentParser) -> None:
 	group_processors = find_argument_group(program, 'processors')
 	if group_processors:
-		group_processors.add_argument('--deep-swapper-model', help = wording.get('help.deep_swapper_model'), default = config.get_str_value('processors', 'deep_swapper_model', 'iperov/elon_musk_224'), choices = processors_choices.deep_swapper_models)
-		group_processors.add_argument('--deep-swapper-morph', help = wording.get('help.deep_swapper_morph'), type = int, default = config.get_int_value('processors', 'deep_swapper_morph', '100'), choices = processors_choices.deep_swapper_morph_range, metavar = create_int_metavar(processors_choices.deep_swapper_morph_range))
+		group_processors.add_argument('--deep-swapper-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'deep_swapper_model', 'iperov/elon_musk_224'), choices = deep_swapper_choices.deep_swapper_models)
+		group_processors.add_argument('--deep-swapper-morph', help = translator.get('help.morph', __package__), type = int, default = config.get_int_value('processors', 'deep_swapper_morph', '100'), choices = deep_swapper_choices.deep_swapper_morph_range, metavar = create_int_metavar(deep_swapper_choices.deep_swapper_morph_range))
 		facefusion.jobs.job_store.register_step_keys([ 'deep_swapper_model', 'deep_swapper_morph' ])
 
 
@@ -279,10 +287,18 @@ def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
 	apply_state_item('deep_swapper_morph', args.get('deep_swapper_morph'))
 
 
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
 def pre_check() -> bool:
 	model_hash_set = get_model_options().get('hashes')
 	model_source_set = get_model_options().get('sources')
 
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
 	if model_hash_set and model_source_set:
 		return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 	return True
@@ -290,28 +306,28 @@ def pre_check() -> bool:
 
 def pre_process(mode : ProcessMode) -> bool:
 	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
 		return False
 	return True
 
 
 def post_process() -> None:
 	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
 	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
 		clear_inference_pool()
+
 	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
 
 
 def swap_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
@@ -319,7 +335,7 @@ def swap_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFram
 	model_size = get_model_size()
 	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
 	crop_vision_frame_raw = crop_vision_frame.copy()
-	box_mask = create_static_box_mask(crop_vision_frame.shape[:2][::-1], state_manager.get_item('face_mask_blur'), state_manager.get_item('face_mask_padding'))
+	box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), state_manager.get_item('face_mask_padding'))
 	crop_masks =\
 	[
 		box_mask
@@ -336,6 +352,11 @@ def swap_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFram
 	crop_vision_frame = conditional_match_frame_color(crop_vision_frame_raw, crop_vision_frame)
 	crop_masks.append(prepare_crop_mask(crop_source_mask, crop_target_mask))
 
+	if 'area' in state_manager.get_item('face_mask_types'):
+		face_landmark_68 = cv2.transform(target_face.landmark_set.get('68').reshape(1, -1, 2), affine_matrix).reshape(-1, 2)
+		area_mask = create_area_mask(crop_vision_frame, face_landmark_68, state_manager.get_item('face_mask_areas'))
+		crop_masks.append(area_mask)
+
 	if 'region' in state_manager.get_item('face_mask_types'):
 		region_mask = create_region_mask(crop_vision_frame, state_manager.get_item('face_mask_regions'))
 		crop_masks.append(region_mask)
@@ -395,56 +416,19 @@ def prepare_crop_mask(crop_source_mask : Mask, crop_target_mask : Mask) -> Mask:
 	return crop_mask
 
 
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	return swap_face(target_face, temp_vision_frame)
+def process_frame(inputs : DeepSwapperInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
 
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
 
-def process_frame(inputs : DeepSwapperInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = swap_face(target_face, temp_vision_frame)
 
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = swap_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = swap_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = swap_face(similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_path : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_path : str, target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(None, temp_frame_paths, process_frames)
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/deep_swapper/locales.py

@@ -0,0 +1,18 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for swapping the face',
+			'morph': 'morph between source face and target faces'
+		},
+		'uis':
+		{
+			'model_dropdown': 'DEEP SWAPPER MODEL',
+			'morph_slider': 'DEEP SWAPPER MORPH'
+		}
+	}
+}

facefusion/processors/modules/deep_swapper/types.py

@@ -0,0 +1,18 @@
+from typing import Any, List, TypeAlias, TypedDict
+
+from numpy.typing import NDArray
+
+from facefusion.types import Mask, VisionFrame
+
+DeepSwapperInputs = TypedDict('DeepSwapperInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+DeepSwapperModel : TypeAlias = str
+
+DeepSwapperMorph : TypeAlias = NDArray[Any]

facefusion/processors/modules/expression_restorer.py

@@ -1,297 +0,0 @@
-from argparse import ArgumentParser
-from functools import lru_cache
-from typing import List, Tuple
-
-import cv2
-import numpy
-
-import facefusion.jobs.job_manager
-import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import create_int_metavar
-from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.face_analyser import get_many_faces, get_one_face
-from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
-from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.live_portrait import create_rotation, limit_expression
-from facefusion.processors.types import ExpressionRestorerInputs, LivePortraitExpression, LivePortraitFeatureVolume, LivePortraitMotionPoints, LivePortraitPitch, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitYaw
-from facefusion.program_helper import find_argument_group
-from facefusion.thread_helper import conditional_thread_semaphore, thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import read_image, read_static_image, read_video_frame, write_image
-
-
-@lru_cache(maxsize = None)
-def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
-	return\
-	{
-		'live_portrait':
-		{
-			'hashes':
-			{
-				'feature_extractor':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.hash'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.hash')
-				},
-				'motion_extractor':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.hash'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.hash')
-				},
-				'generator':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.hash'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_generator.hash')
-				}
-			},
-			'sources':
-			{
-				'feature_extractor':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.onnx'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.onnx')
-				},
-				'motion_extractor':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.onnx'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.onnx')
-				},
-				'generator':
-				{
-					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.onnx'),
-					'path': resolve_relative_path('../.assets/models/live_portrait_generator.onnx')
-				}
-			},
-			'template': 'arcface_128',
-			'size': (512, 512)
-		}
-	}
-
-
-def get_inference_pool() -> InferencePool:
-	model_names = [ state_manager.get_item('expression_restorer_model') ]
-	model_source_set = get_model_options().get('sources')
-
-	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
-
-
-def clear_inference_pool() -> None:
-	model_names = [ state_manager.get_item('expression_restorer_model') ]
-	inference_manager.clear_inference_pool(__name__, model_names)
-
-
-def get_model_options() -> ModelOptions:
-	model_name = state_manager.get_item('expression_restorer_model')
-	return create_static_model_set('full').get(model_name)
-
-
-def register_args(program : ArgumentParser) -> None:
-	group_processors = find_argument_group(program, 'processors')
-	if group_processors:
-		group_processors.add_argument('--expression-restorer-model', help = wording.get('help.expression_restorer_model'), default = config.get_str_value('processors', 'expression_restorer_model', 'live_portrait'), choices = processors_choices.expression_restorer_models)
-		group_processors.add_argument('--expression-restorer-factor', help = wording.get('help.expression_restorer_factor'), type = int, default = config.get_int_value('processors', 'expression_restorer_factor', '80'), choices = processors_choices.expression_restorer_factor_range, metavar = create_int_metavar(processors_choices.expression_restorer_factor_range))
-		facefusion.jobs.job_store.register_step_keys([ 'expression_restorer_model', 'expression_restorer_factor' ])
-
-
-def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
-	apply_state_item('expression_restorer_model', args.get('expression_restorer_model'))
-	apply_state_item('expression_restorer_factor', args.get('expression_restorer_factor'))
-
-
-def pre_check() -> bool:
-	model_hash_set = get_model_options().get('hashes')
-	model_source_set = get_model_options().get('sources')
-
-	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
-
-
-def pre_process(mode : ProcessMode) -> bool:
-	if mode == 'stream':
-		logger.error(wording.get('stream_not_supported') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
-		return False
-	return True
-
-
-def post_process() -> None:
-	read_static_image.cache_clear()
-	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
-		clear_inference_pool()
-	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
-
-
-def restore_expression(source_vision_frame : VisionFrame, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	model_template = get_model_options().get('template')
-	model_size = get_model_options().get('size')
-	expression_restorer_factor = float(numpy.interp(float(state_manager.get_item('expression_restorer_factor')), [ 0, 100 ], [ 0, 1.2 ]))
-	source_vision_frame = cv2.resize(source_vision_frame, temp_vision_frame.shape[:2][::-1])
-	source_crop_vision_frame, _ = warp_face_by_face_landmark_5(source_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
-	target_crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
-	box_mask = create_static_box_mask(target_crop_vision_frame.shape[:2][::-1], state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
-	crop_masks =\
-	[
-		box_mask
-	]
-
-	if 'occlusion' in state_manager.get_item('face_mask_types'):
-		occlusion_mask = create_occlusion_mask(target_crop_vision_frame)
-		crop_masks.append(occlusion_mask)
-
-	source_crop_vision_frame = prepare_crop_frame(source_crop_vision_frame)
-	target_crop_vision_frame = prepare_crop_frame(target_crop_vision_frame)
-	target_crop_vision_frame = apply_restore(source_crop_vision_frame, target_crop_vision_frame, expression_restorer_factor)
-	target_crop_vision_frame = normalize_crop_frame(target_crop_vision_frame)
-	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
-	temp_vision_frame = paste_back(temp_vision_frame, target_crop_vision_frame, crop_mask, affine_matrix)
-	return temp_vision_frame
-
-
-def apply_restore(source_crop_vision_frame : VisionFrame, target_crop_vision_frame : VisionFrame, expression_restorer_factor : float) -> VisionFrame:
-	feature_volume = forward_extract_feature(target_crop_vision_frame)
-	source_expression = forward_extract_motion(source_crop_vision_frame)[5]
-	pitch, yaw, roll, scale, translation, target_expression, motion_points = forward_extract_motion(target_crop_vision_frame)
-	rotation = create_rotation(pitch, yaw, roll)
-	source_expression[:, [ 0, 4, 5, 8, 9 ]] = target_expression[:, [ 0, 4, 5, 8, 9 ]]
-	source_expression = source_expression * expression_restorer_factor + target_expression * (1 - expression_restorer_factor)
-	source_expression = limit_expression(source_expression)
-	source_motion_points = scale * (motion_points @ rotation.T + source_expression) + translation
-	target_motion_points = scale * (motion_points @ rotation.T + target_expression) + translation
-	crop_vision_frame = forward_generate_frame(feature_volume, source_motion_points, target_motion_points)
-	return crop_vision_frame
-
-
-def forward_extract_feature(crop_vision_frame : VisionFrame) -> LivePortraitFeatureVolume:
-	feature_extractor = get_inference_pool().get('feature_extractor')
-
-	with conditional_thread_semaphore():
-		feature_volume = feature_extractor.run(None,
-		{
-			'input': crop_vision_frame
-		})[0]
-
-	return feature_volume
-
-
-def forward_extract_motion(crop_vision_frame : VisionFrame) -> Tuple[LivePortraitPitch, LivePortraitYaw, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitExpression, LivePortraitMotionPoints]:
-	motion_extractor = get_inference_pool().get('motion_extractor')
-
-	with conditional_thread_semaphore():
-		pitch, yaw, roll, scale, translation, expression, motion_points = motion_extractor.run(None,
-		{
-			'input': crop_vision_frame
-		})
-
-	return pitch, yaw, roll, scale, translation, expression, motion_points
-
-
-def forward_generate_frame(feature_volume : LivePortraitFeatureVolume, source_motion_points : LivePortraitMotionPoints, target_motion_points : LivePortraitMotionPoints) -> VisionFrame:
-	generator = get_inference_pool().get('generator')
-
-	with thread_semaphore():
-		crop_vision_frame = generator.run(None,
-		{
-			'feature_volume': feature_volume,
-			'source': source_motion_points,
-			'target': target_motion_points
-		})[0][0]
-
-	return crop_vision_frame
-
-
-def prepare_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
-	model_size = get_model_options().get('size')
-	prepare_size = (model_size[0] // 2, model_size[1] // 2)
-	crop_vision_frame = cv2.resize(crop_vision_frame, prepare_size, interpolation = cv2.INTER_AREA)
-	crop_vision_frame = crop_vision_frame[:, :, ::-1] / 255.0
-	crop_vision_frame = numpy.expand_dims(crop_vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
-	return crop_vision_frame
-
-
-def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
-	crop_vision_frame = crop_vision_frame.transpose(1, 2, 0).clip(0, 1)
-	crop_vision_frame = crop_vision_frame * 255.0
-	crop_vision_frame = crop_vision_frame.astype(numpy.uint8)[:, :, ::-1]
-	return crop_vision_frame
-
-
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	pass
-
-
-def process_frame(inputs : ExpressionRestorerInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	source_vision_frame = inputs.get('source_vision_frame')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
-
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = restore_expression(source_vision_frame, target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = restore_expression(source_vision_frame, target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = restore_expression(source_vision_frame, similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_path : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		frame_number = queue_payload.get('frame_number')
-		if state_manager.get_item('trim_frame_start'):
-			frame_number += state_manager.get_item('trim_frame_start')
-		source_vision_frame = read_video_frame(state_manager.get_item('target_path'), frame_number)
-		target_vision_path = queue_payload.get('frame_path')
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'source_vision_frame': source_vision_frame,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_path : str, target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	source_vision_frame = read_static_image(state_manager.get_item('target_path'))
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'source_vision_frame': source_vision_frame,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(None, temp_frame_paths, process_frames)

facefusion/processors/modules/expression_restorer/choices.py

@@ -0,0 +1,10 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_int_range
+from facefusion.processors.modules.expression_restorer.types import ExpressionRestorerArea, ExpressionRestorerModel
+
+expression_restorer_models : List[ExpressionRestorerModel] = list(get_args(ExpressionRestorerModel))
+
+expression_restorer_areas : List[ExpressionRestorerArea] = list(get_args(ExpressionRestorerArea))
+
+expression_restorer_factor_range : Sequence[int] = create_int_range(0, 100, 1)

facefusion/processors/modules/expression_restorer/core.py

@@ -0,0 +1,280 @@
+from argparse import ArgumentParser
+from functools import lru_cache
+from types import ModuleType
+from typing import List, Tuple
+
+import cv2
+import numpy
+
+import facefusion.jobs.job_manager
+import facefusion.jobs.job_store
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_int_metavar, get_middle
+from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
+from facefusion.face_creator import scale_face
+from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
+from facefusion.face_masker import create_box_mask, create_occlusion_mask
+from facefusion.face_selector import select_faces
+from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
+from facefusion.processors.live_portrait import create_rotation, limit_expression
+from facefusion.processors.modules.expression_restorer import choices as expression_restorer_choices
+from facefusion.processors.modules.expression_restorer.types import ExpressionRestorerInputs
+from facefusion.processors.types import LivePortraitExpression, LivePortraitFeatureVolume, LivePortraitMotionPoints, LivePortraitPitch, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitYaw, ProcessorOutputs
+from facefusion.program_helper import find_argument_group
+from facefusion.thread_helper import conditional_thread_semaphore, thread_semaphore
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_video_frame
+
+
+@lru_cache()
+def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
+	return\
+	{
+		'live_portrait':
+		{
+			'__metadata__':
+			{
+				'vendor': 'KwaiVGI',
+				'license': 'MIT',
+				'year': 2024
+			},
+			'hashes':
+			{
+				'feature_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.hash')
+				},
+				'motion_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.hash')
+				},
+				'generator':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.hash'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_generator.hash')
+				}
+			},
+			'sources':
+			{
+				'feature_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_feature_extractor.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_feature_extractor.onnx')
+				},
+				'motion_extractor':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_motion_extractor.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_motion_extractor.onnx')
+				},
+				'generator':
+				{
+					'url': resolve_download_url('models-3.0.0', 'live_portrait_generator.onnx'),
+					'path': resolve_relative_path('../.assets/models/live_portrait_generator.onnx')
+				}
+			},
+			'template': 'arcface_128',
+			'size': (512, 512)
+		}
+	}
+
+
+def get_inference_pool() -> InferencePool:
+	model_names = [ state_manager.get_item('expression_restorer_model') ]
+	model_source_set = get_model_options().get('sources')
+
+	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
+
+
+def clear_inference_pool() -> None:
+	model_names = [ state_manager.get_item('expression_restorer_model') ]
+	inference_manager.clear_inference_pool(__name__, model_names)
+
+
+def get_model_options() -> ModelOptions:
+	model_name = state_manager.get_item('expression_restorer_model')
+	return create_static_model_set('full').get(model_name)
+
+
+def register_args(program : ArgumentParser) -> None:
+	group_processors = find_argument_group(program, 'processors')
+	if group_processors:
+		group_processors.add_argument('--expression-restorer-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'expression_restorer_model', 'live_portrait'), choices = expression_restorer_choices.expression_restorer_models)
+		group_processors.add_argument('--expression-restorer-factor', help = translator.get('help.factor', __package__), type = int, default = config.get_int_value('processors', 'expression_restorer_factor', '80'), choices = expression_restorer_choices.expression_restorer_factor_range, metavar = create_int_metavar(expression_restorer_choices.expression_restorer_factor_range))
+		group_processors.add_argument('--expression-restorer-areas', help = translator.get('help.areas', __package__).format(choices = ', '.join(expression_restorer_choices.expression_restorer_areas)), default = config.get_str_list('processors', 'expression_restorer_areas', ' '.join(expression_restorer_choices.expression_restorer_areas)), choices = expression_restorer_choices.expression_restorer_areas, nargs = '+', metavar = 'EXPRESSION_RESTORER_AREAS')
+		facefusion.jobs.job_store.register_step_keys([ 'expression_restorer_model', 'expression_restorer_factor', 'expression_restorer_areas' ])
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('expression_restorer_model', args.get('expression_restorer_model'))
+	apply_state_item('expression_restorer_factor', args.get('expression_restorer_factor'))
+	apply_state_item('expression_restorer_areas', args.get('expression_restorer_areas'))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
+def pre_check() -> bool:
+	model_hash_set = get_model_options().get('hashes')
+	model_source_set = get_model_options().get('sources')
+
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
+	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
+
+
+def pre_process(mode : ProcessMode) -> bool:
+	if mode == 'stream':
+		logger.error(translator.get('stream_not_supported') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
+		return False
+	return True
+
+
+def post_process() -> None:
+	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
+	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
+		clear_inference_pool()
+
+	if state_manager.get_item('video_memory_strategy') == 'strict':
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
+
+
+def restore_expression(target_face : Face, target_vision_frame : VisionFrame, temp_vision_frame : VisionFrame) -> VisionFrame:
+	model_template = get_model_options().get('template')
+	model_size = get_model_options().get('size')
+	expression_restorer_factor = float(numpy.interp(float(state_manager.get_item('expression_restorer_factor')), [ 0, 100 ], [ 0, 1.2 ]))
+	target_crop_vision_frame, _ = warp_face_by_face_landmark_5(target_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
+	temp_crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
+	box_mask = create_box_mask(temp_crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+	crop_masks =\
+	[
+		box_mask
+	]
+
+	if 'occlusion' in state_manager.get_item('face_mask_types'):
+		occlusion_mask = create_occlusion_mask(temp_crop_vision_frame)
+		crop_masks.append(occlusion_mask)
+
+	target_crop_vision_frame = prepare_crop_frame(target_crop_vision_frame)
+	temp_crop_vision_frame = prepare_crop_frame(temp_crop_vision_frame)
+	temp_crop_vision_frame = apply_restore(target_crop_vision_frame, temp_crop_vision_frame, expression_restorer_factor)
+	temp_crop_vision_frame = normalize_crop_frame(temp_crop_vision_frame)
+	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+	paste_vision_frame = paste_back(temp_vision_frame, temp_crop_vision_frame, crop_mask, affine_matrix)
+	return paste_vision_frame
+
+
+def apply_restore(target_crop_vision_frame : VisionFrame, temp_crop_vision_frame : VisionFrame, expression_restorer_factor : float) -> VisionFrame:
+	feature_volume = forward_extract_feature(temp_crop_vision_frame)
+	target_expression = forward_extract_motion(target_crop_vision_frame)[5]
+	pitch, yaw, roll, scale, translation, temp_expression, motion_points = forward_extract_motion(temp_crop_vision_frame)
+	rotation = create_rotation(pitch, yaw, roll)
+	target_expression = restrict_expression_areas(temp_expression, target_expression)
+	target_expression = target_expression * expression_restorer_factor + temp_expression * (1 - expression_restorer_factor)
+	target_expression = limit_expression(target_expression)
+	target_motion_points = scale * (motion_points @ rotation.T + target_expression) + translation
+	temp_motion_points = scale * (motion_points @ rotation.T + temp_expression) + translation
+	crop_vision_frame = forward_generate_frame(feature_volume, target_motion_points, temp_motion_points)
+	return crop_vision_frame
+
+
+def restrict_expression_areas(temp_expression : LivePortraitExpression, target_expression : LivePortraitExpression) -> LivePortraitExpression:
+	expression_restorer_areas = state_manager.get_item('expression_restorer_areas')
+
+	if 'upper-face' not in expression_restorer_areas:
+		target_expression[:, [ 1, 2, 6, 10, 11, 12, 13, 15, 16 ]] = temp_expression[:, [ 1, 2, 6, 10, 11, 12, 13, 15, 16 ]]
+
+	if 'lower-face' not in expression_restorer_areas:
+		target_expression[:, [ 3, 7, 14, 17, 18, 19, 20 ]] = temp_expression[:, [ 3, 7, 14, 17, 18, 19, 20 ]]
+
+	target_expression[:, [ 0, 4, 5, 8, 9 ]] = temp_expression[:, [ 0, 4, 5, 8, 9 ]]
+	return target_expression
+
+
+def forward_extract_feature(crop_vision_frame : VisionFrame) -> LivePortraitFeatureVolume:
+	feature_extractor = get_inference_pool().get('feature_extractor')
+
+	with conditional_thread_semaphore():
+		feature_volume = feature_extractor.run(None,
+		{
+			'input': crop_vision_frame
+		})[0]
+
+	return feature_volume
+
+
+def forward_extract_motion(crop_vision_frame : VisionFrame) -> Tuple[LivePortraitPitch, LivePortraitYaw, LivePortraitRoll, LivePortraitScale, LivePortraitTranslation, LivePortraitExpression, LivePortraitMotionPoints]:
+	motion_extractor = get_inference_pool().get('motion_extractor')
+
+	with conditional_thread_semaphore():
+		pitch, yaw, roll, scale, translation, expression, motion_points = motion_extractor.run(None,
+		{
+			'input': crop_vision_frame
+		})
+
+	return pitch, yaw, roll, scale, translation, expression, motion_points
+
+
+def forward_generate_frame(feature_volume : LivePortraitFeatureVolume, target_motion_points : LivePortraitMotionPoints, temp_motion_points : LivePortraitMotionPoints) -> VisionFrame:
+	generator = get_inference_pool().get('generator')
+
+	with thread_semaphore():
+		crop_vision_frame = generator.run(None,
+		{
+			'feature_volume': feature_volume,
+			'source': target_motion_points,
+			'target': temp_motion_points
+		})[0][0]
+
+	return crop_vision_frame
+
+
+def prepare_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
+	model_size = get_model_options().get('size')
+	prepare_size = (model_size[0] // 2, model_size[1] // 2)
+	crop_vision_frame = cv2.resize(crop_vision_frame, prepare_size, interpolation = cv2.INTER_AREA)
+	crop_vision_frame = crop_vision_frame[:, :, ::-1] / 255.0
+	crop_vision_frame = numpy.expand_dims(crop_vision_frame.transpose(2, 0, 1), axis = 0).astype(numpy.float32)
+	return crop_vision_frame
+
+
+def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
+	crop_vision_frame = crop_vision_frame.transpose(1, 2, 0).clip(0, 1)
+	crop_vision_frame = crop_vision_frame * 255.0
+	crop_vision_frame = crop_vision_frame.astype(numpy.uint8)[:, :, ::-1]
+	return crop_vision_frame
+
+
+def process_frame(inputs : ExpressionRestorerInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
+
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
+
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = restore_expression(target_face, target_vision_frame, temp_vision_frame)
+
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/expression_restorer/locales.py

@@ -0,0 +1,20 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for restoring the expression',
+			'factor': 'restore factor of expression from the target face',
+			'areas': 'choose the items used for the expression areas (choices: {choices})'
+		},
+		'uis':
+		{
+			'model_dropdown': 'EXPRESSION RESTORER MODEL',
+			'factor_slider': 'EXPRESSION RESTORER FACTOR',
+			'areas_checkbox_group': 'EXPRESSION RESTORER AREAS'
+		}
+	}
+}

facefusion/processors/modules/expression_restorer/types.py

@@ -0,0 +1,16 @@
+from typing import List, Literal, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+ExpressionRestorerInputs = TypedDict('ExpressionRestorerInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+ExpressionRestorerModel = Literal['live_portrait']
+
+ExpressionRestorerArea = Literal['upper-face', 'lower-face']

facefusion/processors/modules/face_debugger.py

@@ -1,222 +0,0 @@
-from argparse import ArgumentParser
-from typing import List
-
-import cv2
-import numpy
-
-import facefusion.jobs.job_manager
-import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, process_manager, state_manager, wording
-from facefusion.face_analyser import get_many_faces, get_one_face
-from facefusion.face_helper import warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_region_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
-from facefusion.filesystem import in_directory, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.types import FaceDebuggerInputs
-from facefusion.program_helper import find_argument_group
-from facefusion.types import ApplyStateItem, Args, Face, InferencePool, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import read_image, read_static_image, write_image
-
-
-def get_inference_pool() -> InferencePool:
-	pass
-
-
-def clear_inference_pool() -> None:
-	pass
-
-
-def register_args(program : ArgumentParser) -> None:
-	group_processors = find_argument_group(program, 'processors')
-	if group_processors:
-		group_processors.add_argument('--face-debugger-items', help = wording.get('help.face_debugger_items').format(choices = ', '.join(processors_choices.face_debugger_items)), default = config.get_str_list('processors', 'face_debugger_items', 'face-landmark-5/68 face-mask'), choices = processors_choices.face_debugger_items, nargs = '+', metavar = 'FACE_DEBUGGER_ITEMS')
-		facefusion.jobs.job_store.register_step_keys([ 'face_debugger_items' ])
-
-
-def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
-	apply_state_item('face_debugger_items', args.get('face_debugger_items'))
-
-
-def pre_check() -> bool:
-	return True
-
-
-def pre_process(mode : ProcessMode) -> bool:
-	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
-		return False
-	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
-		return False
-	return True
-
-
-def post_process() -> None:
-	read_static_image.cache_clear()
-	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
-
-
-def debug_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	primary_color = (0, 0, 255)
-	primary_light_color = (100, 100, 255)
-	secondary_color = (0, 255, 0)
-	tertiary_color = (255, 255, 0)
-	bounding_box = target_face.bounding_box.astype(numpy.int32)
-	temp_vision_frame = temp_vision_frame.copy()
-	has_face_landmark_5_fallback = numpy.array_equal(target_face.landmark_set.get('5'), target_face.landmark_set.get('5/68'))
-	has_face_landmark_68_fallback = numpy.array_equal(target_face.landmark_set.get('68'), target_face.landmark_set.get('68/5'))
-	face_debugger_items = state_manager.get_item('face_debugger_items')
-
-	if 'bounding-box' in face_debugger_items:
-		x1, y1, x2, y2 = bounding_box
-		cv2.rectangle(temp_vision_frame, (x1, y1), (x2, y2), primary_color, 2)
-
-		if target_face.angle == 0:
-			cv2.line(temp_vision_frame, (x1, y1), (x2, y1), primary_light_color, 3)
-		elif target_face.angle == 180:
-			cv2.line(temp_vision_frame, (x1, y2), (x2, y2), primary_light_color, 3)
-		elif target_face.angle == 90:
-			cv2.line(temp_vision_frame, (x2, y1), (x2, y2), primary_light_color, 3)
-		elif target_face.angle == 270:
-			cv2.line(temp_vision_frame, (x1, y1), (x1, y2), primary_light_color, 3)
-
-	if 'face-mask' in face_debugger_items:
-		crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), 'arcface_128', (512, 512))
-		inverse_matrix = cv2.invertAffineTransform(affine_matrix)
-		temp_size = temp_vision_frame.shape[:2][::-1]
-		crop_masks = []
-
-		if 'box' in state_manager.get_item('face_mask_types'):
-			box_mask = create_static_box_mask(crop_vision_frame.shape[:2][::-1], 0, state_manager.get_item('face_mask_padding'))
-			crop_masks.append(box_mask)
-
-		if 'occlusion' in state_manager.get_item('face_mask_types'):
-			occlusion_mask = create_occlusion_mask(crop_vision_frame)
-			crop_masks.append(occlusion_mask)
-
-		if 'region' in state_manager.get_item('face_mask_types'):
-			region_mask = create_region_mask(crop_vision_frame, state_manager.get_item('face_mask_regions'))
-			crop_masks.append(region_mask)
-
-		crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
-		crop_mask = (crop_mask * 255).astype(numpy.uint8)
-		inverse_vision_frame = cv2.warpAffine(crop_mask, inverse_matrix, temp_size)
-		inverse_vision_frame = cv2.threshold(inverse_vision_frame, 100, 255, cv2.THRESH_BINARY)[1]
-		inverse_vision_frame[inverse_vision_frame > 0] = 255 #type:ignore[operator]
-		inverse_contours = cv2.findContours(inverse_vision_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE)[0]
-		cv2.drawContours(temp_vision_frame, inverse_contours, -1, tertiary_color if has_face_landmark_5_fallback else secondary_color, 2)
-
-	if 'face-landmark-5' in face_debugger_items and numpy.any(target_face.landmark_set.get('5')):
-		face_landmark_5 = target_face.landmark_set.get('5').astype(numpy.int32)
-		for index in range(face_landmark_5.shape[0]):
-			cv2.circle(temp_vision_frame, (face_landmark_5[index][0], face_landmark_5[index][1]), 3, primary_color, -1)
-
-	if 'face-landmark-5/68' in face_debugger_items and numpy.any(target_face.landmark_set.get('5/68')):
-		face_landmark_5_68 = target_face.landmark_set.get('5/68').astype(numpy.int32)
-		for index in range(face_landmark_5_68.shape[0]):
-			cv2.circle(temp_vision_frame, (face_landmark_5_68[index][0], face_landmark_5_68[index][1]), 3, tertiary_color if has_face_landmark_5_fallback else secondary_color, -1)
-
-	if 'face-landmark-68' in face_debugger_items and numpy.any(target_face.landmark_set.get('68')):
-		face_landmark_68 = target_face.landmark_set.get('68').astype(numpy.int32)
-		for index in range(face_landmark_68.shape[0]):
-			cv2.circle(temp_vision_frame, (face_landmark_68[index][0], face_landmark_68[index][1]), 3, tertiary_color if has_face_landmark_68_fallback else secondary_color, -1)
-
-	if 'face-landmark-68/5' in face_debugger_items and numpy.any(target_face.landmark_set.get('68')):
-		face_landmark_68 = target_face.landmark_set.get('68/5').astype(numpy.int32)
-		for index in range(face_landmark_68.shape[0]):
-			cv2.circle(temp_vision_frame, (face_landmark_68[index][0], face_landmark_68[index][1]), 3, tertiary_color, -1)
-
-	if bounding_box[3] - bounding_box[1] > 50 and bounding_box[2] - bounding_box[0] > 50:
-		top = bounding_box[1]
-		left = bounding_box[0] - 20
-
-		if 'face-detector-score' in face_debugger_items:
-			face_score_text = str(round(target_face.score_set.get('detector'), 2))
-			top = top + 20
-			cv2.putText(temp_vision_frame, face_score_text, (left, top), cv2.FONT_HERSHEY_SIMPLEX, 0.5, primary_color, 2)
-
-		if 'face-landmarker-score' in face_debugger_items:
-			face_score_text = str(round(target_face.score_set.get('landmarker'), 2))
-			top = top + 20
-			cv2.putText(temp_vision_frame, face_score_text, (left, top), cv2.FONT_HERSHEY_SIMPLEX, 0.5, tertiary_color if has_face_landmark_5_fallback else secondary_color, 2)
-
-		if 'age' in face_debugger_items:
-			face_age_text = str(target_face.age.start) + '-' + str(target_face.age.stop)
-			top = top + 20
-			cv2.putText(temp_vision_frame, face_age_text, (left, top), cv2.FONT_HERSHEY_SIMPLEX, 0.5, primary_color, 2)
-
-		if 'gender' in face_debugger_items:
-			face_gender_text = target_face.gender
-			top = top + 20
-			cv2.putText(temp_vision_frame, face_gender_text, (left, top), cv2.FONT_HERSHEY_SIMPLEX, 0.5, primary_color, 2)
-
-		if 'race' in face_debugger_items:
-			face_race_text = target_face.race
-			top = top + 20
-			cv2.putText(temp_vision_frame, face_race_text, (left, top), cv2.FONT_HERSHEY_SIMPLEX, 0.5, primary_color, 2)
-
-	return temp_vision_frame
-
-
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	pass
-
-
-def process_frame(inputs : FaceDebuggerInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
-
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = debug_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = debug_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = debug_face(similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_paths : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_paths : List[str], target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(source_paths, temp_frame_paths, process_frames)

facefusion/processors/modules/face_debugger/choices.py

@@ -0,0 +1,5 @@
+from typing import List, get_args
+
+from facefusion.processors.modules.face_debugger.types import FaceDebuggerItem
+
+face_debugger_items : List[FaceDebuggerItem] = list(get_args(FaceDebuggerItem))

facefusion/processors/modules/face_debugger/core.py

@@ -0,0 +1,270 @@
+from argparse import ArgumentParser
+from types import ModuleType
+from typing import List
+
+import cv2
+import numpy
+
+import facefusion.jobs.job_manager
+import facefusion.jobs.job_store
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, state_manager, translator, video_manager
+from facefusion.common_helper import get_middle
+from facefusion.face_creator import scale_face
+from facefusion.face_helper import warp_face_by_face_landmark_5
+from facefusion.face_masker import create_area_mask, create_box_mask, create_occlusion_mask, create_region_mask
+from facefusion.face_selector import select_faces
+from facefusion.filesystem import in_directory, is_image, is_video, same_file_extension
+from facefusion.processors.modules.face_debugger import choices as face_debugger_choices
+from facefusion.processors.modules.face_debugger.types import FaceDebuggerInputs
+from facefusion.processors.types import ProcessorOutputs
+from facefusion.program_helper import find_argument_group
+from facefusion.types import ApplyStateItem, Args, Face, InferencePool, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_video_frame
+
+
+def get_inference_pool() -> InferencePool:
+	pass
+
+
+def clear_inference_pool() -> None:
+	pass
+
+
+def register_args(program : ArgumentParser) -> None:
+	group_processors = find_argument_group(program, 'processors')
+	if group_processors:
+		group_processors.add_argument('--face-debugger-items', help = translator.get('help.items', __package__).format(choices = ', '.join(face_debugger_choices.face_debugger_items)), default = config.get_str_list('processors', 'face_debugger_items', 'face-landmark-5/68 face-mask'), choices = face_debugger_choices.face_debugger_items, nargs = '+', metavar = 'FACE_DEBUGGER_ITEMS')
+		facefusion.jobs.job_store.register_step_keys([ 'face_debugger_items' ])
+
+
+def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
+	apply_state_item('face_debugger_items', args.get('face_debugger_items'))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
+def pre_check() -> bool:
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+	return True
+
+
+def pre_process(mode : ProcessMode) -> bool:
+	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
+		return False
+	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
+		return False
+	return True
+
+
+def post_process() -> None:
+	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
+	if state_manager.get_item('video_memory_strategy') == 'strict':
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
+
+
+def debug_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	face_debugger_items = state_manager.get_item('face_debugger_items')
+
+	if 'bounding-box' in face_debugger_items:
+		temp_vision_frame = draw_bounding_box(target_face, temp_vision_frame)
+
+	if 'face-mask' in face_debugger_items:
+		temp_vision_frame = draw_face_mask(target_face, temp_vision_frame)
+
+	if 'face-landmark-5' in face_debugger_items:
+		temp_vision_frame = draw_face_landmark_5(target_face, temp_vision_frame)
+
+	if 'face-landmark-5/68' in face_debugger_items:
+		temp_vision_frame = draw_face_landmark_5_68(target_face, temp_vision_frame)
+
+	if 'face-landmark-68' in face_debugger_items:
+		temp_vision_frame = draw_face_landmark_68(target_face, temp_vision_frame)
+
+	if 'face-landmark-68/5' in face_debugger_items:
+		temp_vision_frame = draw_face_landmark_68_5(target_face, temp_vision_frame)
+
+	return temp_vision_frame
+
+
+def draw_bounding_box(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	bounding_box = target_face.bounding_box.astype(numpy.int32)
+	x1, y1, x2, y2 = bounding_box
+	box_color = 0, 0, 255
+	border_scale = calculate_scale(temp_vision_frame)
+	border_color = 100, 100, 255
+
+	cv2.rectangle(temp_vision_frame, (x1, y1), (x2, y2), box_color, border_scale)
+
+	if target_face.angle == 0:
+		cv2.line(temp_vision_frame, (x1, y1), (x2, y1), border_color, border_scale + 1)
+	if target_face.angle == 180:
+		cv2.line(temp_vision_frame, (x1, y2), (x2, y2), border_color, border_scale + 1)
+	if target_face.angle == 90:
+		cv2.line(temp_vision_frame, (x2, y1), (x2, y2), border_color, border_scale + 1)
+	if target_face.angle == 270:
+		cv2.line(temp_vision_frame, (x1, y1), (x1, y2), border_color, border_scale + 1)
+
+	return temp_vision_frame
+
+
+def draw_face_mask(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	crop_masks = []
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	face_landmark_5 = target_face.landmark_set.get('5')
+	face_landmark_68 = target_face.landmark_set.get('68')
+	face_landmark_5_68 = target_face.landmark_set.get('5/68')
+	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5_68, 'arcface_128', (512, 512))
+	inverse_matrix = cv2.invertAffineTransform(affine_matrix)
+	temp_size = temp_vision_frame.shape[:2][::-1]
+	mask_scale = calculate_scale(temp_vision_frame)
+	mask_color = 0, 255, 0
+
+	if numpy.array_equal(face_landmark_5, face_landmark_5_68):
+		mask_color = 255, 255, 0
+
+	if target_face.origin == 'refill':
+		mask_color = 0, 165, 255
+
+	if 'box' in state_manager.get_item('face_mask_types'):
+		box_mask = create_box_mask(crop_vision_frame, 0, state_manager.get_item('face_mask_padding'))
+		crop_masks.append(box_mask)
+
+	if 'occlusion' in state_manager.get_item('face_mask_types'):
+		occlusion_mask = create_occlusion_mask(crop_vision_frame)
+		crop_masks.append(occlusion_mask)
+
+	if 'area' in state_manager.get_item('face_mask_types'):
+		face_landmark_68 = cv2.transform(face_landmark_68.reshape(1, -1, 2), affine_matrix).reshape(-1, 2)
+		area_mask = create_area_mask(crop_vision_frame, face_landmark_68, state_manager.get_item('face_mask_areas'))
+		crop_masks.append(area_mask)
+
+	if 'region' in state_manager.get_item('face_mask_types'):
+		region_mask = create_region_mask(crop_vision_frame, state_manager.get_item('face_mask_regions'))
+		crop_masks.append(region_mask)
+
+	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
+	crop_mask = (crop_mask * 255).astype(numpy.uint8)
+	inverse_vision_frame = cv2.warpAffine(crop_mask, inverse_matrix, temp_size)
+	inverse_vision_frame = cv2.threshold(inverse_vision_frame, 100, 255, cv2.THRESH_BINARY)[1]
+	inverse_contours, _ = cv2.findContours(inverse_vision_frame, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE)
+	cv2.drawContours(temp_vision_frame, inverse_contours, -1, mask_color, mask_scale)
+
+	return temp_vision_frame
+
+
+def draw_face_landmark_5(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	face_landmark_5 = target_face.landmark_set.get('5')
+	point_scale = calculate_scale(temp_vision_frame)
+	point_color = 0, 0, 255
+
+	if target_face.origin == 'refill':
+		point_color = 0, 165, 255
+
+	if numpy.any(face_landmark_5):
+		face_landmark_5 = face_landmark_5.astype(numpy.int32)
+
+		for point in face_landmark_5:
+			cv2.circle(temp_vision_frame, tuple(point), point_scale, point_color, -1)
+
+	return temp_vision_frame
+
+
+def draw_face_landmark_5_68(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	face_landmark_5 = target_face.landmark_set.get('5')
+	face_landmark_5_68 = target_face.landmark_set.get('5/68')
+	point_scale = calculate_scale(temp_vision_frame)
+	point_color = 0, 255, 0
+
+	if numpy.array_equal(face_landmark_5, face_landmark_5_68):
+		point_color = 255, 255, 0
+
+	if target_face.origin == 'refill':
+		point_color = 0, 165, 255
+
+	if numpy.any(face_landmark_5_68):
+		face_landmark_5_68 = face_landmark_5_68.astype(numpy.int32)
+
+		for point in face_landmark_5_68:
+			cv2.circle(temp_vision_frame, tuple(point), point_scale, point_color, -1)
+
+	return temp_vision_frame
+
+
+def draw_face_landmark_68(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	face_landmark_68 = target_face.landmark_set.get('68')
+	face_landmark_68_5 = target_face.landmark_set.get('68/5')
+	point_scale = calculate_scale(temp_vision_frame)
+	point_color = 0, 255, 0
+
+	if numpy.array_equal(face_landmark_68, face_landmark_68_5):
+		point_color = 255, 255, 0
+
+	if target_face.origin == 'refill':
+		point_color = 0, 165, 255
+
+	if numpy.any(face_landmark_68):
+		face_landmark_68 = face_landmark_68.astype(numpy.int32)
+
+		for point in face_landmark_68:
+			cv2.circle(temp_vision_frame, tuple(point), point_scale, point_color, -1)
+
+	return temp_vision_frame
+
+
+def draw_face_landmark_68_5(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+	temp_vision_frame = numpy.ascontiguousarray(temp_vision_frame)
+	face_landmark_68_5 = target_face.landmark_set.get('68/5')
+	point_scale = calculate_scale(temp_vision_frame)
+	point_color = 255, 255, 0
+
+	if target_face.origin == 'refill':
+		point_color = 0, 165, 255
+
+	if numpy.any(face_landmark_68_5):
+		face_landmark_68_5 = face_landmark_68_5.astype(numpy.int32)
+
+		for point in face_landmark_68_5:
+			cv2.circle(temp_vision_frame, tuple(point), point_scale, point_color, -1)
+
+	return temp_vision_frame
+
+
+def calculate_scale(temp_vision_frame : VisionFrame) -> int:
+	frame_height, _ = temp_vision_frame.shape[:2]
+	frame_scale = round(frame_height / 270)
+	return max(1, min(10, frame_scale))
+
+
+def process_frame(inputs : FaceDebuggerInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
+
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
+
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = debug_face(target_face, temp_vision_frame)
+
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/face_debugger/locales.py

@@ -0,0 +1,16 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'items': 'load a single or multiple processors (choices: {choices})'
+		},
+		'uis':
+		{
+			'items_checkbox_group': 'FACE DEBUGGER ITEMS'
+		}
+	}
+}

facefusion/processors/modules/face_debugger/types.py

@@ -0,0 +1,14 @@
+from typing import List, Literal, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+FaceDebuggerInputs = TypedDict('FaceDebuggerInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+FaceDebuggerItem = Literal['bounding-box', 'face-landmark-5', 'face-landmark-5/68', 'face-landmark-68', 'face-landmark-68/5', 'face-mask']

facefusion/processors/modules/face_editor/choices.py

@@ -0,0 +1,21 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_float_range
+from facefusion.processors.modules.face_editor.types import FaceEditorModel
+
+face_editor_models : List[FaceEditorModel] = list(get_args(FaceEditorModel))
+
+face_editor_eyebrow_direction_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_eye_gaze_horizontal_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_eye_gaze_vertical_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_eye_open_ratio_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_lip_open_ratio_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_grim_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_pout_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_purse_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_smile_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_position_horizontal_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_mouth_position_vertical_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_head_pitch_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_head_yaw_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)
+face_editor_head_roll_range : Sequence[float] = create_float_range(-1.0, 1.0, 0.05)

facefusion/processors/modules/face_editor/core.py

@@ -1,5 +1,6 @@
 from argparse import ArgumentParser
 from functools import lru_cache
+from types import ModuleType
 from typing import List, Tuple
 
 import cv2
@@ -7,31 +8,36 @@ import numpy
 
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import create_float_metavar
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_float_metavar, get_middle
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.face_analyser import get_many_faces, get_one_face
+from facefusion.face_creator import scale_face
 from facefusion.face_helper import paste_back, scale_face_landmark_5, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
+from facefusion.face_masker import create_box_mask
+from facefusion.face_selector import select_faces
 from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.live_portrait import create_rotation, limit_euler_angles, limit_expression
-from facefusion.processors.types import FaceEditorInputs, LivePortraitExpression, LivePortraitFeatureVolume, LivePortraitMotionPoints, LivePortraitPitch, LivePortraitRoll, LivePortraitRotation, LivePortraitScale, LivePortraitTranslation, LivePortraitYaw
+from facefusion.processors.live_portrait import create_rotation, limit_angle, limit_expression
+from facefusion.processors.modules.face_editor import choices as face_editor_choices
+from facefusion.processors.modules.face_editor.types import FaceEditorInputs
+from facefusion.processors.types import LivePortraitExpression, LivePortraitFeatureVolume, LivePortraitMotionPoints, LivePortraitPitch, LivePortraitRoll, LivePortraitRotation, LivePortraitScale, LivePortraitTranslation, LivePortraitYaw, ProcessorOutputs
 from facefusion.program_helper import find_argument_group
 from facefusion.thread_helper import conditional_thread_semaphore, thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, FaceLandmark68, InferencePool, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import read_image, read_static_image, write_image
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, FaceLandmark68, InferencePool, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_video_frame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'live_portrait':
 		{
+			'__metadata__':
+			{
+				'vendor': 'KwaiVGI',
+				'license': 'MIT',
+				'year': 2024
+			},
 			'hashes':
 			{
 				'feature_extractor':
@@ -124,21 +130,21 @@ def get_model_options() -> ModelOptions:
 def register_args(program : ArgumentParser) -> None:
 	group_processors = find_argument_group(program, 'processors')
 	if group_processors:
-		group_processors.add_argument('--face-editor-model', help = wording.get('help.face_editor_model'), default = config.get_str_value('processors', 'face_editor_model', 'live_portrait'), choices = processors_choices.face_editor_models)
-		group_processors.add_argument('--face-editor-eyebrow-direction', help = wording.get('help.face_editor_eyebrow_direction'), type = float, default = config.get_float_value('processors', 'face_editor_eyebrow_direction', '0'), choices = processors_choices.face_editor_eyebrow_direction_range, metavar = create_float_metavar(processors_choices.face_editor_eyebrow_direction_range))
-		group_processors.add_argument('--face-editor-eye-gaze-horizontal', help = wording.get('help.face_editor_eye_gaze_horizontal'), type = float, default = config.get_float_value('processors', 'face_editor_eye_gaze_horizontal', '0'), choices = processors_choices.face_editor_eye_gaze_horizontal_range, metavar = create_float_metavar(processors_choices.face_editor_eye_gaze_horizontal_range))
-		group_processors.add_argument('--face-editor-eye-gaze-vertical', help = wording.get('help.face_editor_eye_gaze_vertical'), type = float, default = config.get_float_value('processors', 'face_editor_eye_gaze_vertical', '0'), choices = processors_choices.face_editor_eye_gaze_vertical_range, metavar = create_float_metavar(processors_choices.face_editor_eye_gaze_vertical_range))
-		group_processors.add_argument('--face-editor-eye-open-ratio', help = wording.get('help.face_editor_eye_open_ratio'), type = float, default = config.get_float_value('processors', 'face_editor_eye_open_ratio', '0'), choices = processors_choices.face_editor_eye_open_ratio_range, metavar = create_float_metavar(processors_choices.face_editor_eye_open_ratio_range))
-		group_processors.add_argument('--face-editor-lip-open-ratio', help = wording.get('help.face_editor_lip_open_ratio'), type = float, default = config.get_float_value('processors', 'face_editor_lip_open_ratio', '0'), choices = processors_choices.face_editor_lip_open_ratio_range, metavar = create_float_metavar(processors_choices.face_editor_lip_open_ratio_range))
-		group_processors.add_argument('--face-editor-mouth-grim', help = wording.get('help.face_editor_mouth_grim'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_grim', '0'), choices = processors_choices.face_editor_mouth_grim_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_grim_range))
-		group_processors.add_argument('--face-editor-mouth-pout', help = wording.get('help.face_editor_mouth_pout'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_pout', '0'), choices = processors_choices.face_editor_mouth_pout_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_pout_range))
-		group_processors.add_argument('--face-editor-mouth-purse', help = wording.get('help.face_editor_mouth_purse'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_purse', '0'), choices = processors_choices.face_editor_mouth_purse_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_purse_range))
-		group_processors.add_argument('--face-editor-mouth-smile', help = wording.get('help.face_editor_mouth_smile'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_smile', '0'), choices = processors_choices.face_editor_mouth_smile_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_smile_range))
-		group_processors.add_argument('--face-editor-mouth-position-horizontal', help = wording.get('help.face_editor_mouth_position_horizontal'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_position_horizontal', '0'), choices = processors_choices.face_editor_mouth_position_horizontal_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_position_horizontal_range))
-		group_processors.add_argument('--face-editor-mouth-position-vertical', help = wording.get('help.face_editor_mouth_position_vertical'), type = float, default = config.get_float_value('processors', 'face_editor_mouth_position_vertical', '0'), choices = processors_choices.face_editor_mouth_position_vertical_range, metavar = create_float_metavar(processors_choices.face_editor_mouth_position_vertical_range))
-		group_processors.add_argument('--face-editor-head-pitch', help = wording.get('help.face_editor_head_pitch'), type = float, default = config.get_float_value('processors', 'face_editor_head_pitch', '0'), choices = processors_choices.face_editor_head_pitch_range, metavar = create_float_metavar(processors_choices.face_editor_head_pitch_range))
-		group_processors.add_argument('--face-editor-head-yaw', help = wording.get('help.face_editor_head_yaw'), type = float, default = config.get_float_value('processors', 'face_editor_head_yaw', '0'), choices = processors_choices.face_editor_head_yaw_range, metavar = create_float_metavar(processors_choices.face_editor_head_yaw_range))
-		group_processors.add_argument('--face-editor-head-roll', help = wording.get('help.face_editor_head_roll'), type = float, default = config.get_float_value('processors', 'face_editor_head_roll', '0'), choices = processors_choices.face_editor_head_roll_range, metavar = create_float_metavar(processors_choices.face_editor_head_roll_range))
+		group_processors.add_argument('--face-editor-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'face_editor_model', 'live_portrait'), choices = face_editor_choices.face_editor_models)
+		group_processors.add_argument('--face-editor-eyebrow-direction', help = translator.get('help.eyebrow_direction', __package__), type = float, default = config.get_float_value('processors', 'face_editor_eyebrow_direction', '0'), choices = face_editor_choices.face_editor_eyebrow_direction_range, metavar = create_float_metavar(face_editor_choices.face_editor_eyebrow_direction_range))
+		group_processors.add_argument('--face-editor-eye-gaze-horizontal', help = translator.get('help.eye_gaze_horizontal', __package__), type = float, default = config.get_float_value('processors', 'face_editor_eye_gaze_horizontal', '0'), choices = face_editor_choices.face_editor_eye_gaze_horizontal_range, metavar = create_float_metavar(face_editor_choices.face_editor_eye_gaze_horizontal_range))
+		group_processors.add_argument('--face-editor-eye-gaze-vertical', help = translator.get('help.eye_gaze_vertical', __package__), type = float, default = config.get_float_value('processors', 'face_editor_eye_gaze_vertical', '0'), choices = face_editor_choices.face_editor_eye_gaze_vertical_range, metavar = create_float_metavar(face_editor_choices.face_editor_eye_gaze_vertical_range))
+		group_processors.add_argument('--face-editor-eye-open-ratio', help = translator.get('help.eye_open_ratio', __package__), type = float, default = config.get_float_value('processors', 'face_editor_eye_open_ratio', '0'), choices = face_editor_choices.face_editor_eye_open_ratio_range, metavar = create_float_metavar(face_editor_choices.face_editor_eye_open_ratio_range))
+		group_processors.add_argument('--face-editor-lip-open-ratio', help = translator.get('help.lip_open_ratio', __package__), type = float, default = config.get_float_value('processors', 'face_editor_lip_open_ratio', '0'), choices = face_editor_choices.face_editor_lip_open_ratio_range, metavar = create_float_metavar(face_editor_choices.face_editor_lip_open_ratio_range))
+		group_processors.add_argument('--face-editor-mouth-grim', help = translator.get('help.mouth_grim', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_grim', '0'), choices = face_editor_choices.face_editor_mouth_grim_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_grim_range))
+		group_processors.add_argument('--face-editor-mouth-pout', help = translator.get('help.mouth_pout', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_pout', '0'), choices = face_editor_choices.face_editor_mouth_pout_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_pout_range))
+		group_processors.add_argument('--face-editor-mouth-purse', help = translator.get('help.mouth_purse', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_purse', '0'), choices = face_editor_choices.face_editor_mouth_purse_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_purse_range))
+		group_processors.add_argument('--face-editor-mouth-smile', help = translator.get('help.mouth_smile', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_smile', '0'), choices = face_editor_choices.face_editor_mouth_smile_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_smile_range))
+		group_processors.add_argument('--face-editor-mouth-position-horizontal', help = translator.get('help.mouth_position_horizontal', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_position_horizontal', '0'), choices = face_editor_choices.face_editor_mouth_position_horizontal_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_position_horizontal_range))
+		group_processors.add_argument('--face-editor-mouth-position-vertical', help = translator.get('help.mouth_position_vertical', __package__), type = float, default = config.get_float_value('processors', 'face_editor_mouth_position_vertical', '0'), choices = face_editor_choices.face_editor_mouth_position_vertical_range, metavar = create_float_metavar(face_editor_choices.face_editor_mouth_position_vertical_range))
+		group_processors.add_argument('--face-editor-head-pitch', help = translator.get('help.head_pitch', __package__), type = float, default = config.get_float_value('processors', 'face_editor_head_pitch', '0'), choices = face_editor_choices.face_editor_head_pitch_range, metavar = create_float_metavar(face_editor_choices.face_editor_head_pitch_range))
+		group_processors.add_argument('--face-editor-head-yaw', help = translator.get('help.head_yaw', __package__), type = float, default = config.get_float_value('processors', 'face_editor_head_yaw', '0'), choices = face_editor_choices.face_editor_head_yaw_range, metavar = create_float_metavar(face_editor_choices.face_editor_head_yaw_range))
+		group_processors.add_argument('--face-editor-head-roll', help = translator.get('help.head_roll', __package__), type = float, default = config.get_float_value('processors', 'face_editor_head_roll', '0'), choices = face_editor_choices.face_editor_head_roll_range, metavar = create_float_metavar(face_editor_choices.face_editor_head_roll_range))
 		facefusion.jobs.job_store.register_step_keys([ 'face_editor_model', 'face_editor_eyebrow_direction', 'face_editor_eye_gaze_horizontal', 'face_editor_eye_gaze_vertical', 'face_editor_eye_open_ratio', 'face_editor_lip_open_ratio', 'face_editor_mouth_grim', 'face_editor_mouth_pout', 'face_editor_mouth_purse', 'face_editor_mouth_smile', 'face_editor_mouth_position_horizontal', 'face_editor_mouth_position_vertical', 'face_editor_head_pitch', 'face_editor_head_yaw', 'face_editor_head_roll' ])
 
 
@@ -160,37 +166,45 @@ def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
 	apply_state_item('face_editor_head_roll', args.get('face_editor_head_roll'))
 
 
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
 def pre_check() -> bool:
 	model_hash_set = get_model_options().get('hashes')
 	model_source_set = get_model_options().get('sources')
 
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
 
 def pre_process(mode : ProcessMode) -> bool:
 	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
 		return False
 	return True
 
 
 def post_process() -> None:
 	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
 	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
 		clear_inference_pool()
+
 	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
 
 
 def edit_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
@@ -198,12 +212,12 @@ def edit_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFram
 	model_size = get_model_options().get('size')
 	face_landmark_5 = scale_face_landmark_5(target_face.landmark_set.get('5/68'), 1.5)
 	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, face_landmark_5, model_template, model_size)
-	box_mask = create_static_box_mask(crop_vision_frame.shape[:2][::-1], state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+	box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
 	crop_vision_frame = prepare_crop_frame(crop_vision_frame)
 	crop_vision_frame = apply_edit(crop_vision_frame, target_face.landmark_set.get('68'))
 	crop_vision_frame = normalize_crop_frame(crop_vision_frame)
-	temp_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, box_mask, affine_matrix)
-	return temp_vision_frame
+	paste_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, box_mask, affine_matrix)
+	return paste_vision_frame
 
 
 def apply_edit(crop_vision_frame : VisionFrame, face_landmark_68 : FaceLandmark68) -> VisionFrame:
@@ -341,8 +355,8 @@ def edit_eye_gaze(expression : LivePortraitExpression) -> LivePortraitExpression
 
 def edit_eye_open(motion_points : LivePortraitMotionPoints, face_landmark_68 : FaceLandmark68) -> LivePortraitMotionPoints:
 	face_editor_eye_open_ratio = state_manager.get_item('face_editor_eye_open_ratio')
-	left_eye_ratio = calc_distance_ratio(face_landmark_68, 37, 40, 39, 36)
-	right_eye_ratio = calc_distance_ratio(face_landmark_68, 43, 46, 45, 42)
+	left_eye_ratio = calculate_distance_ratio(face_landmark_68, 37, 40, 39, 36)
+	right_eye_ratio = calculate_distance_ratio(face_landmark_68, 43, 46, 45, 42)
 
 	if face_editor_eye_open_ratio < 0:
 		eye_motion_points = numpy.concatenate([ motion_points.ravel(), [ left_eye_ratio, right_eye_ratio, 0.0 ] ])
@@ -356,7 +370,7 @@ def edit_eye_open(motion_points : LivePortraitMotionPoints, face_landmark_68 : F
 
 def edit_lip_open(motion_points : LivePortraitMotionPoints, face_landmark_68 : FaceLandmark68) -> LivePortraitMotionPoints:
 	face_editor_lip_open_ratio = state_manager.get_item('face_editor_lip_open_ratio')
-	lip_ratio = calc_distance_ratio(face_landmark_68, 62, 66, 54, 48)
+	lip_ratio = calculate_distance_ratio(face_landmark_68, 62, 66, 54, 48)
 
 	if face_editor_lip_open_ratio < 0:
 		lip_motion_points = numpy.concatenate([ motion_points.ravel(), [ lip_ratio, 0.0 ] ])
@@ -449,12 +463,12 @@ def edit_head_rotation(pitch : LivePortraitPitch, yaw : LivePortraitYaw, roll :
 	edit_pitch = pitch + float(numpy.interp(face_editor_head_pitch, [ -1, 1 ], [ 20, -20 ]))
 	edit_yaw = yaw + float(numpy.interp(face_editor_head_yaw, [ -1, 1 ], [ 60, -60 ]))
 	edit_roll = roll + float(numpy.interp(face_editor_head_roll, [ -1, 1 ], [ -15, 15 ]))
-	edit_pitch, edit_yaw, edit_roll = limit_euler_angles(pitch, yaw, roll, edit_pitch, edit_yaw, edit_roll)
+	edit_pitch, edit_yaw, edit_roll = limit_angle(pitch, yaw, roll, edit_pitch, edit_yaw, edit_roll)
 	rotation = create_rotation(edit_pitch, edit_yaw, edit_roll)
 	return rotation
 
 
-def calc_distance_ratio(face_landmark_68 : FaceLandmark68, top_index : int, bottom_index : int, left_index : int, right_index : int) -> float:
+def calculate_distance_ratio(face_landmark_68 : FaceLandmark68, top_index : int, bottom_index : int, left_index : int, right_index : int) -> float:
 	vertical_direction = face_landmark_68[top_index] - face_landmark_68[bottom_index]
 	horizontal_direction = face_landmark_68[left_index] - face_landmark_68[right_index]
 	distance_ratio = float(numpy.linalg.norm(vertical_direction) / (numpy.linalg.norm(horizontal_direction) + 1e-6))
@@ -477,56 +491,19 @@ def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
 	return crop_vision_frame
 
 
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	pass
+def process_frame(inputs : FaceEditorInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
 
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
 
-def process_frame(inputs : FaceEditorInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = edit_face(target_face, temp_vision_frame)
 
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = edit_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = edit_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = edit_face(similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_path : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_path : str, target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(None, temp_frame_paths, process_frames)
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/face_editor/locales.py

@@ -0,0 +1,44 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for editing the face',
+			'eyebrow_direction': 'specify the eyebrow direction',
+			'eye_gaze_horizontal': 'specify the horizontal eye gaze',
+			'eye_gaze_vertical': 'specify the vertical eye gaze',
+			'eye_open_ratio': 'specify the ratio of eye opening',
+			'lip_open_ratio': 'specify the ratio of lip opening',
+			'mouth_grim': 'specify the mouth grim',
+			'mouth_pout': 'specify the mouth pout',
+			'mouth_purse': 'specify the mouth purse',
+			'mouth_smile': 'specify the mouth smile',
+			'mouth_position_horizontal': 'specify the horizontal mouth position',
+			'mouth_position_vertical': 'specify the vertical mouth position',
+			'head_pitch': 'specify the head pitch',
+			'head_yaw': 'specify the head yaw',
+			'head_roll': 'specify the head roll'
+		},
+		'uis':
+		{
+			'eyebrow_direction_slider': 'FACE EDITOR EYEBROW DIRECTION',
+			'eye_gaze_horizontal_slider': 'FACE EDITOR EYE GAZE HORIZONTAL',
+			'eye_gaze_vertical_slider': 'FACE EDITOR EYE GAZE VERTICAL',
+			'eye_open_ratio_slider': 'FACE EDITOR EYE OPEN RATIO',
+			'head_pitch_slider': 'FACE EDITOR HEAD PITCH',
+			'head_roll_slider': 'FACE EDITOR HEAD ROLL',
+			'head_yaw_slider': 'FACE EDITOR HEAD YAW',
+			'lip_open_ratio_slider': 'FACE EDITOR LIP OPEN RATIO',
+			'model_dropdown': 'FACE EDITOR MODEL',
+			'mouth_grim_slider': 'FACE EDITOR MOUTH GRIM',
+			'mouth_position_horizontal_slider': 'FACE EDITOR MOUTH POSITION HORIZONTAL',
+			'mouth_position_vertical_slider': 'FACE EDITOR MOUTH POSITION VERTICAL',
+			'mouth_pout_slider': 'FACE EDITOR MOUTH POUT',
+			'mouth_purse_slider': 'FACE EDITOR MOUTH PURSE',
+			'mouth_smile_slider': 'FACE EDITOR MOUTH SMILE'
+		}
+	}
+}

facefusion/processors/modules/face_editor/types.py

@@ -0,0 +1,14 @@
+from typing import List, Literal, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+FaceEditorInputs = TypedDict('FaceEditorInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+FaceEditorModel = Literal['live_portrait']

facefusion/processors/modules/face_enhancer/choices.py

@@ -0,0 +1,10 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_float_range, create_int_range
+from facefusion.processors.modules.face_enhancer.types import FaceEnhancerModel
+
+face_enhancer_models : List[FaceEnhancerModel] = list(get_args(FaceEnhancerModel))
+
+face_enhancer_blend_range : Sequence[int] = create_int_range(0, 100, 1)
+
+face_enhancer_weight_range : Sequence[float] = create_float_range(0.0, 1.0, 0.05)

facefusion/processors/modules/face_enhancer/core.py

@@ -1,36 +1,41 @@
 from argparse import ArgumentParser
 from functools import lru_cache
+from types import ModuleType
 from typing import List
 
-import cv2
 import numpy
 
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import create_float_metavar, create_int_metavar
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import create_float_metavar, create_int_metavar, get_middle
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
-from facefusion.face_analyser import get_many_faces, get_one_face
+from facefusion.face_creator import scale_face
 from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces
-from facefusion.face_store import get_reference_faces
+from facefusion.face_masker import create_box_mask, create_occlusion_mask
+from facefusion.face_selector import select_faces
 from facefusion.filesystem import in_directory, is_image, is_video, resolve_relative_path, same_file_extension
-from facefusion.processors import choices as processors_choices
-from facefusion.processors.types import FaceEnhancerInputs, FaceEnhancerWeight
+from facefusion.processors.modules.face_enhancer import choices as face_enhancer_choices
+from facefusion.processors.modules.face_enhancer.types import FaceEnhancerInputs, FaceEnhancerWeight
+from facefusion.processors.types import ProcessorOutputs
 from facefusion.program_helper import find_argument_group
 from facefusion.thread_helper import thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import read_image, read_static_image, write_image
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import blend_frame, read_static_image, read_static_video_frame
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'codeformer':
 		{
+			'__metadata__':
+			{
+				'vendor': 'sczhou',
+				'license': 'S-Lab-1.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -52,6 +57,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gfpgan_1.2':
 		{
+			'__metadata__':
+			{
+				'vendor': 'TencentARC',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -73,6 +84,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gfpgan_1.3':
 		{
+			'__metadata__':
+			{
+				'vendor': 'TencentARC',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -94,6 +111,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gfpgan_1.4':
 		{
+			'__metadata__':
+			{
+				'vendor': 'TencentARC',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -115,6 +138,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gpen_bfr_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yangxy',
+				'license': 'Non-Commercial',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -136,6 +165,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gpen_bfr_512':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yangxy',
+				'license': 'Non-Commercial',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -157,6 +192,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gpen_bfr_1024':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yangxy',
+				'license': 'Non-Commercial',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -178,6 +219,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'gpen_bfr_2048':
 		{
+			'__metadata__':
+			{
+				'vendor': 'yangxy',
+				'license': 'Non-Commercial',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -199,6 +246,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'restoreformer_plus_plus':
 		{
+			'__metadata__':
+			{
+				'vendor': 'wzhouxiff',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_enhancer':
@@ -241,9 +294,9 @@ def get_model_options() -> ModelOptions:
 def register_args(program : ArgumentParser) -> None:
 	group_processors = find_argument_group(program, 'processors')
 	if group_processors:
-		group_processors.add_argument('--face-enhancer-model', help = wording.get('help.face_enhancer_model'), default = config.get_str_value('processors', 'face_enhancer_model', 'gfpgan_1.4'), choices = processors_choices.face_enhancer_models)
-		group_processors.add_argument('--face-enhancer-blend', help = wording.get('help.face_enhancer_blend'), type = int, default = config.get_int_value('processors', 'face_enhancer_blend', '80'), choices = processors_choices.face_enhancer_blend_range, metavar = create_int_metavar(processors_choices.face_enhancer_blend_range))
-		group_processors.add_argument('--face-enhancer-weight', help = wording.get('help.face_enhancer_weight'), type = float, default = config.get_float_value('processors', 'face_enhancer_weight', '1.0'), choices = processors_choices.face_enhancer_weight_range, metavar = create_float_metavar(processors_choices.face_enhancer_weight_range))
+		group_processors.add_argument('--face-enhancer-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'face_enhancer_model', 'gfpgan_1.4'), choices = face_enhancer_choices.face_enhancer_models)
+		group_processors.add_argument('--face-enhancer-blend', help = translator.get('help.blend', __package__), type = int, default = config.get_int_value('processors', 'face_enhancer_blend', '80'), choices = face_enhancer_choices.face_enhancer_blend_range, metavar = create_int_metavar(face_enhancer_choices.face_enhancer_blend_range))
+		group_processors.add_argument('--face-enhancer-weight', help = translator.get('help.weight', __package__), type = float, default = config.get_float_value('processors', 'face_enhancer_weight', '0.5'), choices = face_enhancer_choices.face_enhancer_weight_range, metavar = create_float_metavar(face_enhancer_choices.face_enhancer_weight_range))
 		facefusion.jobs.job_store.register_step_keys([ 'face_enhancer_model', 'face_enhancer_blend', 'face_enhancer_weight' ])
 
 
@@ -253,44 +306,52 @@ def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
 	apply_state_item('face_enhancer_weight', args.get('face_enhancer_weight'))
 
 
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
+
+
 def pre_check() -> bool:
 	model_hash_set = get_model_options().get('hashes')
 	model_source_set = get_model_options().get('sources')
 
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
 
 def pre_process(mode : ProcessMode) -> bool:
 	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
 		return False
 	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
 		return False
 	return True
 
 
 def post_process() -> None:
 	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
 	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
 		clear_inference_pool()
+
 	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
 
 
 def enhance_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
 	model_template = get_model_options().get('template')
 	model_size = get_model_options().get('size')
 	crop_vision_frame, affine_matrix = warp_face_by_face_landmark_5(temp_vision_frame, target_face.landmark_set.get('5/68'), model_template, model_size)
-	box_mask = create_static_box_mask(crop_vision_frame.shape[:2][::-1], state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
+	box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), (0, 0, 0, 0))
 	crop_masks =\
 	[
 		box_mask
@@ -306,7 +367,7 @@ def enhance_face(target_face : Face, temp_vision_frame : VisionFrame) -> VisionF
 	crop_vision_frame = normalize_crop_frame(crop_vision_frame)
 	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
 	paste_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, crop_mask, affine_matrix)
-	temp_vision_frame = blend_frame(temp_vision_frame, paste_vision_frame)
+	temp_vision_frame = blend_paste_frame(temp_vision_frame, paste_vision_frame)
 	return temp_vision_frame
 
 
@@ -352,62 +413,25 @@ def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
 	return crop_vision_frame
 
 
-def blend_frame(temp_vision_frame : VisionFrame, paste_vision_frame : VisionFrame) -> VisionFrame:
+def blend_paste_frame(temp_vision_frame : VisionFrame, paste_vision_frame : VisionFrame) -> VisionFrame:
 	face_enhancer_blend = 1 - (state_manager.get_item('face_enhancer_blend') / 100)
-	temp_vision_frame = cv2.addWeighted(temp_vision_frame, face_enhancer_blend, paste_vision_frame, 1 - face_enhancer_blend, 0)
+	temp_vision_frame = blend_frame(temp_vision_frame, paste_vision_frame, 1 - face_enhancer_blend)
 	return temp_vision_frame
 
 
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	return enhance_face(target_face, temp_vision_frame)
+def process_frame(inputs : FaceEnhancerInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
 
+	target_vision_frame = get_middle(target_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
 
-def process_frame(inputs : FaceEnhancerInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
+	if target_faces:
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = enhance_face(target_face, temp_vision_frame)
 
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = enhance_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = enhance_face(target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = enhance_face(similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_path : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
-
-
-def process_image(source_path : str, target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
-
-
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(None, temp_frame_paths, process_frames)
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/face_enhancer/locales.py

@@ -0,0 +1,20 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for enhancing the face',
+			'blend': 'blend the enhanced into the previous face',
+			'weight': 'specify the degree of weight applied to the face'
+		},
+		'uis':
+		{
+			'blend_slider': 'FACE ENHANCER BLEND',
+			'model_dropdown': 'FACE ENHANCER MODEL',
+			'weight_slider': 'FACE ENHANCER WEIGHT'
+		}
+	}
+}

facefusion/processors/modules/face_enhancer/types.py

@@ -0,0 +1,18 @@
+from typing import Any, List, Literal, TypeAlias, TypedDict
+
+from numpy.typing import NDArray
+
+from facefusion.types import Mask, VisionFrame
+
+FaceEnhancerInputs = TypedDict('FaceEnhancerInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+FaceEnhancerModel = Literal['codeformer', 'gfpgan_1.2', 'gfpgan_1.3', 'gfpgan_1.4', 'gpen_bfr_256', 'gpen_bfr_512', 'gpen_bfr_1024', 'gpen_bfr_2048', 'restoreformer_plus_plus']
+
+FaceEnhancerWeight : TypeAlias = NDArray[Any]

facefusion/processors/modules/face_swapper/choices.py

@@ -0,0 +1,26 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_float_range
+from facefusion.processors.modules.face_swapper.types import FaceSwapperModel, FaceSwapperSet, FaceSwapperWeight
+
+
+face_swapper_set : FaceSwapperSet =\
+{
+	'blendswap_256': [ '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
+	'ghost_1_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'ghost_2_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'ghost_3_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'hififace_unofficial_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'hyperswap_1a_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'hyperswap_1b_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'hyperswap_1c_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'inswapper_128': [ '128x128', '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
+	'inswapper_128_fp16': [ '128x128', '256x256', '384x384', '512x512', '768x768', '1024x1024' ],
+	'simswap_256': [ '256x256', '512x512', '768x768', '1024x1024' ],
+	'simswap_unofficial_512': [ '512x512', '768x768', '1024x1024' ],
+	'uniface_256': [ '256x256', '512x512', '768x768', '1024x1024' ]
+}
+
+face_swapper_models : List[FaceSwapperModel] = list(get_args(FaceSwapperModel))
+
+face_swapper_weight_range : Sequence[FaceSwapperWeight] = create_float_range(0.0, 1.0, 0.05)

facefusion/processors/modules/face_swapper/core.py

@@ -1,39 +1,46 @@
 from argparse import ArgumentParser
 from functools import lru_cache
-from typing import List, Tuple
+from types import ModuleType
+from typing import List, Optional, Tuple
 
+import cv2
 import numpy
 
 import facefusion.choices
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
-import facefusion.processors.core as processors
-from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, wording
-from facefusion.common_helper import get_first
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, state_manager, translator, video_manager
+from facefusion.common_helper import get_first, get_middle, is_macos
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
 from facefusion.execution import has_execution_provider
-from facefusion.face_analyser import get_average_face, get_many_faces, get_one_face
+from facefusion.face_creator import average_face_identity, get_one_face, get_static_faces, scale_face
 from facefusion.face_helper import paste_back, warp_face_by_face_landmark_5
-from facefusion.face_masker import create_occlusion_mask, create_region_mask, create_static_box_mask
-from facefusion.face_selector import find_similar_faces, sort_and_filter_faces, sort_faces_by_order
-from facefusion.face_store import get_reference_faces
+from facefusion.face_masker import create_area_mask, create_box_mask, create_occlusion_mask, create_region_mask
+from facefusion.face_selector import select_faces, sort_faces_by_order
 from facefusion.filesystem import filter_image_paths, has_image, in_directory, is_image, is_video, resolve_relative_path, same_file_extension
 from facefusion.model_helper import get_static_model_initializer
-from facefusion.processors import choices as processors_choices
+from facefusion.processors.modules.face_swapper import choices as face_swapper_choices
+from facefusion.processors.modules.face_swapper.types import FaceSwapperInputs
 from facefusion.processors.pixel_boost import explode_pixel_boost, implode_pixel_boost
-from facefusion.processors.types import FaceSwapperInputs
+from facefusion.processors.types import ProcessorOutputs
 from facefusion.program_helper import find_argument_group
 from facefusion.thread_helper import conditional_thread_semaphore
-from facefusion.types import ApplyStateItem, Args, DownloadScope, Embedding, Face, InferencePool, ModelOptions, ModelSet, ProcessMode, QueuePayload, UpdateProgress, VisionFrame
-from facefusion.vision import read_image, read_static_image, read_static_images, unpack_resolution, write_image
+from facefusion.types import ApplyStateItem, Args, DownloadScope, Embedding, Face, InferencePool, InferenceProvider, ModelOptions, ModelSet, ProcessMode, VisionFrame
+from facefusion.vision import read_static_image, read_static_images, read_static_video_frame, unpack_resolution
 
 
-@lru_cache(maxsize = None)
+@lru_cache()
 def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 	return\
 	{
 		'blendswap_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'mapooon',
+				'license': 'Non-Commercial',
+				'year': 2023
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -58,6 +65,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'ghost_1_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'ai-forever',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -67,8 +80,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.hash')
 				}
 			},
 			'sources':
@@ -80,8 +93,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.onnx')
 				}
 			},
 			'type': 'ghost',
@@ -92,6 +105,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'ghost_2_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'ai-forever',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -101,8 +120,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.hash')
 				}
 			},
 			'sources':
@@ -114,8 +133,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.onnx')
 				}
 			},
 			'type': 'ghost',
@@ -126,6 +145,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'ghost_3_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'ai-forever',
+				'license': 'Apache-2.0',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -135,8 +160,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.hash')
 				}
 			},
 			'sources':
@@ -148,8 +173,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_ghost.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_ghost.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_ghost.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_ghost.onnx')
 				}
 			},
 			'type': 'ghost',
@@ -160,6 +185,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'hififace_unofficial_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'GuijiAI',
+				'license': 'Unknown',
+				'year': 2021
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -169,8 +200,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.1.0', 'arcface_converter_hififace.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_hififace.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_hififace.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_hififace.hash')
 				}
 			},
 			'sources':
@@ -182,8 +213,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.1.0', 'arcface_converter_hififace.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_hififace.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_hififace.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_hififace.onnx')
 				}
 			},
 			'type': 'hififace',
@@ -192,8 +223,107 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 			'mean': [ 0.5, 0.5, 0.5 ],
 			'standard_deviation': [ 0.5, 0.5, 0.5 ]
 		},
+		'hyperswap_1a_256':
+		{
+			'__metadata__':
+			{
+				'vendor': 'FaceFusion',
+				'license': 'ResearchRAIL',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'face_swapper':
+				{
+					'url': resolve_download_url('models-3.3.0', 'hyperswap_1a_256.hash'),
+					'path': resolve_relative_path('../.assets/models/hyperswap_1a_256.hash')
+				}
+			},
+			'sources':
+			{
+				'face_swapper':
+				{
+					'url': resolve_download_url('models-3.3.0', 'hyperswap_1a_256.onnx'),
+					'path': resolve_relative_path('../.assets/models/hyperswap_1a_256.onnx')
+				}
+			},
+			'precision': 'fp16',
+			'type': 'hyperswap',
+			'template': 'arcface_128',
+			'size': (256, 256),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
+		},
+		'hyperswap_1b_256':
+		{
+			'__metadata__':
+			{
+				'vendor': 'FaceFusion',
+				'license': 'ResearchRAIL',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'face_swapper':
+				{
+					'url': resolve_download_url('models-3.3.0', 'hyperswap_1b_256.hash'),
+					'path': resolve_relative_path('../.assets/models/hyperswap_1b_256.hash')
+				}
+			},
+			'sources':
+				{
+					'face_swapper':
+					{
+						'url': resolve_download_url('models-3.3.0', 'hyperswap_1b_256.onnx'),
+						'path': resolve_relative_path('../.assets/models/hyperswap_1b_256.onnx')
+					}
+				},
+			'precision': 'fp16',
+			'type': 'hyperswap',
+			'template': 'arcface_128',
+			'size': (256, 256),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
+		},
+		'hyperswap_1c_256':
+		{
+			'__metadata__':
+			{
+				'vendor': 'FaceFusion',
+				'license': 'ResearchRAIL',
+				'year': 2025
+			},
+			'hashes':
+			{
+				'face_swapper':
+				{
+					'url': resolve_download_url('models-3.3.0', 'hyperswap_1c_256.hash'),
+					'path': resolve_relative_path('../.assets/models/hyperswap_1c_256.hash')
+				}
+			},
+			'sources':
+			{
+				'face_swapper':
+				{
+					'url': resolve_download_url('models-3.3.0', 'hyperswap_1c_256.onnx'),
+					'path': resolve_relative_path('../.assets/models/hyperswap_1c_256.onnx')
+				}
+			},
+			'precision': 'fp16',
+			'type': 'hyperswap',
+			'template': 'arcface_128',
+			'size': (256, 256),
+			'mean': [ 0.5, 0.5, 0.5 ],
+			'standard_deviation': [ 0.5, 0.5, 0.5 ]
+		},
 		'inswapper_128':
 		{
+			'__metadata__':
+			{
+				'vendor': 'InsightFace',
+				'license': 'Non-Commercial',
+				'year': 2023
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -218,6 +348,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'inswapper_128_fp16':
 		{
+			'__metadata__':
+			{
+				'vendor': 'InsightFace',
+				'license': 'Non-Commercial',
+				'year': 2023
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -234,6 +370,7 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 					'path': resolve_relative_path('../.assets/models/inswapper_128_fp16.onnx')
 				}
 			},
+			'precision': 'fp16',
 			'type': 'inswapper',
 			'template': 'arcface_128',
 			'size': (128, 128),
@@ -242,6 +379,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'simswap_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'neuralchen',
+				'license': 'Non-Commercial',
+				'year': 2020
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -251,8 +394,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_simswap.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_simswap.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_simswap.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_simswap.hash')
 				}
 			},
 			'sources':
@@ -264,8 +407,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_simswap.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_simswap.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_simswap.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_simswap.onnx')
 				}
 			},
 			'type': 'simswap',
@@ -276,6 +419,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'simswap_unofficial_512':
 		{
+			'__metadata__':
+			{
+				'vendor': 'neuralchen',
+				'license': 'Non-Commercial',
+				'year': 2020
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -285,8 +434,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_simswap.hash'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_simswap.hash')
+					'url': resolve_download_url('models-3.4.0', 'crossface_simswap.hash'),
+					'path': resolve_relative_path('../.assets/models/crossface_simswap.hash')
 				}
 			},
 			'sources':
@@ -298,8 +447,8 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 				},
 				'embedding_converter':
 				{
-					'url': resolve_download_url('models-3.0.0', 'arcface_converter_simswap.onnx'),
-					'path': resolve_relative_path('../.assets/models/arcface_converter_simswap.onnx')
+					'url': resolve_download_url('models-3.4.0', 'crossface_simswap.onnx'),
+					'path': resolve_relative_path('../.assets/models/crossface_simswap.onnx')
 				}
 			},
 			'type': 'simswap',
@@ -310,6 +459,12 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 		},
 		'uniface_256':
 		{
+			'__metadata__':
+			{
+				'vendor': 'xc-csc101',
+				'license': 'Unknown',
+				'year': 2022
+			},
 			'hashes':
 			{
 				'face_swapper':
@@ -336,89 +491,114 @@ def create_static_model_set(download_scope : DownloadScope) -> ModelSet:
 
 
 def get_inference_pool() -> InferencePool:
-	model_names = [ get_model_name() ]
+	model_names = [ state_manager.get_item('face_swapper_model') ]
 	model_source_set = get_model_options().get('sources')
 
 	return inference_manager.get_inference_pool(__name__, model_names, model_source_set)
 
 
 def clear_inference_pool() -> None:
-	model_names = [ get_model_name() ]
+	model_names = [ state_manager.get_item('face_swapper_model') ]
 	inference_manager.clear_inference_pool(__name__, model_names)
 
 
+def adjust_inference_providers() -> List[InferenceProvider]:
+	model_precision = get_model_options().get('precision')
+	model_type = get_model_options().get('type')
+
+	if is_macos() and has_execution_provider('coreml'):
+		if model_type in [ 'ghost', 'uniface' ] or model_precision == 'fp16':
+			return\
+			[
+				(facefusion.choices.execution_provider_set.get('coreml'),
+				{
+					'ModelFormat': 'MLProgram'
+				})
+			]
+
+	return []
+
+
 def get_model_options() -> ModelOptions:
-	model_name = get_model_name()
-	return create_static_model_set('full').get(model_name)
-
-
-def get_model_name() -> str:
 	model_name = state_manager.get_item('face_swapper_model')
-
-	if has_execution_provider('coreml') and model_name == 'inswapper_128_fp16':
-		return 'inswapper_128'
-	return model_name
+	return create_static_model_set('full').get(model_name)
 
 
 def register_args(program : ArgumentParser) -> None:
 	group_processors = find_argument_group(program, 'processors')
 	if group_processors:
-		group_processors.add_argument('--face-swapper-model', help = wording.get('help.face_swapper_model'), default = config.get_str_value('processors', 'face_swapper_model', 'inswapper_128_fp16'), choices = processors_choices.face_swapper_models)
+		group_processors.add_argument('--face-swapper-model', help = translator.get('help.model', __package__), default = config.get_str_value('processors', 'face_swapper_model', 'hyperswap_1a_256'), choices = face_swapper_choices.face_swapper_models)
 		known_args, _ = program.parse_known_args()
-		face_swapper_pixel_boost_choices = processors_choices.face_swapper_set.get(known_args.face_swapper_model)
-		group_processors.add_argument('--face-swapper-pixel-boost', help = wording.get('help.face_swapper_pixel_boost'), default = config.get_str_value('processors', 'face_swapper_pixel_boost', get_first(face_swapper_pixel_boost_choices)), choices = face_swapper_pixel_boost_choices)
-		facefusion.jobs.job_store.register_step_keys([ 'face_swapper_model', 'face_swapper_pixel_boost' ])
+		face_swapper_pixel_boost_choices = face_swapper_choices.face_swapper_set.get(known_args.face_swapper_model)
+		group_processors.add_argument('--face-swapper-pixel-boost', help = translator.get('help.pixel_boost', __package__), default = config.get_str_value('processors', 'face_swapper_pixel_boost', get_first(face_swapper_pixel_boost_choices)), choices = face_swapper_pixel_boost_choices)
+		group_processors.add_argument('--face-swapper-weight', help = translator.get('help.weight', __package__), type = float, default = config.get_float_value('processors', 'face_swapper_weight', '0.5'), choices = face_swapper_choices.face_swapper_weight_range)
+		facefusion.jobs.job_store.register_step_keys([ 'face_swapper_model', 'face_swapper_pixel_boost', 'face_swapper_weight' ])
 
 
 def apply_args(args : Args, apply_state_item : ApplyStateItem) -> None:
 	apply_state_item('face_swapper_model', args.get('face_swapper_model'))
 	apply_state_item('face_swapper_pixel_boost', args.get('face_swapper_pixel_boost'))
+	apply_state_item('face_swapper_weight', args.get('face_swapper_weight'))
+
+
+def get_common_modules() -> List[ModuleType]:
+	return [ content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer ]
 
 
 def pre_check() -> bool:
 	model_hash_set = get_model_options().get('hashes')
 	model_source_set = get_model_options().get('sources')
 
+	for common_module in get_common_modules():
+		if not common_module.pre_check():
+			return False
+
 	return conditional_download_hashes(model_hash_set) and conditional_download_sources(model_source_set)
 
 
 def pre_process(mode : ProcessMode) -> bool:
 	if not has_image(state_manager.get_item('source_paths')):
-		logger.error(wording.get('choose_image_source') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('choose_image_source') + translator.get('exclamation_mark'), __name__)
 		return False
+
 	source_image_paths = filter_image_paths(state_manager.get_item('source_paths'))
-	source_frames = read_static_images(source_image_paths)
-	source_faces = get_many_faces(source_frames)
+	source_vision_frames = read_static_images(source_image_paths)
+	source_faces = get_static_faces(source_vision_frames)
+
 	if not get_one_face(source_faces):
-		logger.error(wording.get('no_source_face_detected') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('no_source_face_detected') + translator.get('exclamation_mark'), __name__)
 		return False
+
 	if mode in [ 'output', 'preview' ] and not is_image(state_manager.get_item('target_path')) and not is_video(state_manager.get_item('target_path')):
-		logger.error(wording.get('choose_image_or_video_target') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('choose_image_or_video_target') + translator.get('exclamation_mark'), __name__)
 		return False
+
 	if mode == 'output' and not in_directory(state_manager.get_item('output_path')):
-		logger.error(wording.get('specify_image_or_video_output') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('specify_image_or_video_output') + translator.get('exclamation_mark'), __name__)
 		return False
+
 	if mode == 'output' and not same_file_extension(state_manager.get_item('target_path'), state_manager.get_item('output_path')):
-		logger.error(wording.get('match_target_and_output_extension') + wording.get('exclamation_mark'), __name__)
+		logger.error(translator.get('match_target_and_output_extension') + translator.get('exclamation_mark'), __name__)
 		return False
+
 	return True
 
 
 def post_process() -> None:
 	read_static_image.cache_clear()
+	read_static_video_frame.cache_clear()
+	video_manager.clear_video_pool()
+
 	if state_manager.get_item('video_memory_strategy') in [ 'strict', 'moderate' ]:
-		clear_inference_pool()
 		get_static_model_initializer.cache_clear()
+		clear_inference_pool()
+
 	if state_manager.get_item('video_memory_strategy') == 'strict':
-		content_analyser.clear_inference_pool()
-		face_classifier.clear_inference_pool()
-		face_detector.clear_inference_pool()
-		face_landmarker.clear_inference_pool()
-		face_masker.clear_inference_pool()
-		face_recognizer.clear_inference_pool()
+		for common_module in get_common_modules():
+			common_module.clear_inference_pool()
 
 
-def swap_face(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
+def swap_face(source_face : Face, target_face : Face, source_vision_frame : VisionFrame, temp_vision_frame : VisionFrame) -> VisionFrame:
 	model_template = get_model_options().get('template')
 	model_size = get_model_options().get('size')
 	pixel_boost_size = unpack_resolution(state_manager.get_item('face_swapper_pixel_boost'))
@@ -428,7 +608,7 @@ def swap_face(source_face : Face, target_face : Face, temp_vision_frame : Vision
 	crop_masks = []
 
 	if 'box' in state_manager.get_item('face_mask_types'):
-		box_mask = create_static_box_mask(crop_vision_frame.shape[:2][::-1], state_manager.get_item('face_mask_blur'), state_manager.get_item('face_mask_padding'))
+		box_mask = create_box_mask(crop_vision_frame, state_manager.get_item('face_mask_blur'), state_manager.get_item('face_mask_padding'))
 		crop_masks.append(box_mask)
 
 	if 'occlusion' in state_manager.get_item('face_mask_types'):
@@ -438,34 +618,38 @@ def swap_face(source_face : Face, target_face : Face, temp_vision_frame : Vision
 	pixel_boost_vision_frames = implode_pixel_boost(crop_vision_frame, pixel_boost_total, model_size)
 	for pixel_boost_vision_frame in pixel_boost_vision_frames:
 		pixel_boost_vision_frame = prepare_crop_frame(pixel_boost_vision_frame)
-		pixel_boost_vision_frame = forward_swap_face(source_face, pixel_boost_vision_frame)
+		pixel_boost_vision_frame = forward_swap_face(source_face, target_face, source_vision_frame, pixel_boost_vision_frame)
 		pixel_boost_vision_frame = normalize_crop_frame(pixel_boost_vision_frame)
 		temp_vision_frames.append(pixel_boost_vision_frame)
 	crop_vision_frame = explode_pixel_boost(temp_vision_frames, pixel_boost_total, model_size, pixel_boost_size)
 
+	if 'area' in state_manager.get_item('face_mask_types'):
+		face_landmark_68 = cv2.transform(target_face.landmark_set.get('68').reshape(1, -1, 2), affine_matrix).reshape(-1, 2)
+		area_mask = create_area_mask(crop_vision_frame, face_landmark_68, state_manager.get_item('face_mask_areas'))
+		crop_masks.append(area_mask)
+
 	if 'region' in state_manager.get_item('face_mask_types'):
 		region_mask = create_region_mask(crop_vision_frame, state_manager.get_item('face_mask_regions'))
 		crop_masks.append(region_mask)
 
 	crop_mask = numpy.minimum.reduce(crop_masks).clip(0, 1)
-	temp_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, crop_mask, affine_matrix)
-	return temp_vision_frame
+	paste_vision_frame = paste_back(temp_vision_frame, crop_vision_frame, crop_mask, affine_matrix)
+	return paste_vision_frame
 
 
-def forward_swap_face(source_face : Face, crop_vision_frame : VisionFrame) -> VisionFrame:
+def forward_swap_face(source_face : Face, target_face : Face, source_vision_frame : VisionFrame, crop_vision_frame : VisionFrame) -> VisionFrame:
 	face_swapper = get_inference_pool().get('face_swapper')
 	model_type = get_model_options().get('type')
 	face_swapper_inputs = {}
 
-	if has_execution_provider('coreml') and model_type in [ 'ghost', 'uniface' ]:
-		face_swapper.set_providers([ facefusion.choices.execution_provider_set.get('cpu') ])
-
 	for face_swapper_input in face_swapper.get_inputs():
 		if face_swapper_input.name == 'source':
 			if model_type in [ 'blendswap', 'uniface' ]:
-				face_swapper_inputs[face_swapper_input.name] = prepare_source_frame(source_face)
+				face_swapper_inputs[face_swapper_input.name] = prepare_source_frame(source_face, source_vision_frame)
 			else:
-				face_swapper_inputs[face_swapper_input.name] = prepare_source_embedding(source_face)
+				source_embedding = prepare_source_embedding(source_face)
+				source_embedding = balance_source_embedding(source_embedding, target_face.embedding)
+				face_swapper_inputs[face_swapper_input.name] = source_embedding
 		if face_swapper_input.name == 'target':
 			face_swapper_inputs[face_swapper_input.name] = crop_vision_frame
 
@@ -475,26 +659,27 @@ def forward_swap_face(source_face : Face, crop_vision_frame : VisionFrame) -> Vi
 	return crop_vision_frame
 
 
-def forward_convert_embedding(embedding : Embedding) -> Embedding:
+def forward_convert_embedding(face_embedding : Embedding) -> Embedding:
 	embedding_converter = get_inference_pool().get('embedding_converter')
 
 	with conditional_thread_semaphore():
-		embedding = embedding_converter.run(None,
+		face_embedding = embedding_converter.run(None,
 		{
-			'input': embedding
+			'input': face_embedding
 		})[0]
 
-	return embedding
+	return face_embedding
 
 
-def prepare_source_frame(source_face : Face) -> VisionFrame:
+def prepare_source_frame(source_face : Face, source_vision_frame : VisionFrame) -> VisionFrame:
 	model_type = get_model_options().get('type')
-	source_vision_frame = read_static_image(get_first(state_manager.get_item('source_paths')))
 
 	if model_type == 'blendswap':
 		source_vision_frame, _ = warp_face_by_face_landmark_5(source_vision_frame, source_face.landmark_set.get('5/68'), 'arcface_112_v2', (112, 112))
+
 	if model_type == 'uniface':
 		source_vision_frame, _ = warp_face_by_face_landmark_5(source_vision_frame, source_face.landmark_set.get('5/68'), 'ffhq_512', (256, 256))
+
 	source_vision_frame = source_vision_frame[:, :, ::-1] / 255.0
 	source_vision_frame = source_vision_frame.transpose(2, 0, 1)
 	source_vision_frame = numpy.expand_dims(source_vision_frame, axis = 0).astype(numpy.float32)
@@ -505,25 +690,47 @@ def prepare_source_embedding(source_face : Face) -> Embedding:
 	model_type = get_model_options().get('type')
 
 	if model_type == 'ghost':
-		source_embedding, _ = convert_embedding(source_face)
+		source_embedding = source_face.embedding.reshape(-1, 512)
+		source_embedding, _ = convert_source_embedding(source_embedding)
 		source_embedding = source_embedding.reshape(1, -1)
-	elif model_type == 'inswapper':
+		return source_embedding
+
+	if model_type == 'hyperswap':
+		source_embedding = source_face.embedding_norm.reshape((1, -1))
+		return source_embedding
+
+	if model_type == 'inswapper':
 		model_path = get_model_options().get('sources').get('face_swapper').get('path')
 		model_initializer = get_static_model_initializer(model_path)
 		source_embedding = source_face.embedding.reshape((1, -1))
 		source_embedding = numpy.dot(source_embedding, model_initializer) / numpy.linalg.norm(source_embedding)
-	else:
-		_, source_normed_embedding = convert_embedding(source_face)
-		source_embedding = source_normed_embedding.reshape(1, -1)
+		return source_embedding
+
+	source_embedding = source_face.embedding.reshape(-1, 512)
+	_, source_embedding_norm = convert_source_embedding(source_embedding)
+	source_embedding = source_embedding_norm.reshape(1, -1)
 	return source_embedding
 
 
-def convert_embedding(source_face : Face) -> Tuple[Embedding, Embedding]:
-	embedding = source_face.embedding.reshape(-1, 512)
-	embedding = forward_convert_embedding(embedding)
-	embedding = embedding.ravel()
-	normed_embedding = embedding / numpy.linalg.norm(embedding)
-	return embedding, normed_embedding
+def balance_source_embedding(source_embedding : Embedding, target_embedding : Embedding) -> Embedding:
+	model_type = get_model_options().get('type')
+	face_swapper_weight = state_manager.get_item('face_swapper_weight')
+	face_swapper_weight = numpy.interp(face_swapper_weight, [ 0, 1 ], [ 0.35, -0.35 ]).astype(numpy.float32)
+
+	if model_type in [ 'hififace', 'hyperswap', 'inswapper', 'simswap' ]:
+		target_embedding = target_embedding / numpy.linalg.norm(target_embedding)
+
+	source_embedding = source_embedding.reshape(1, -1)
+	target_embedding = target_embedding.reshape(1, -1)
+	source_embedding = source_embedding * (1 - face_swapper_weight) + target_embedding * face_swapper_weight
+	return source_embedding
+
+
+def convert_source_embedding(source_embedding : Embedding) -> Tuple[Embedding, Embedding]:
+	source_embedding = forward_convert_embedding(source_embedding)
+	source_embedding = source_embedding.ravel()
+	source_embedding_norm = source_embedding / numpy.linalg.norm(source_embedding)
+	return source_embedding, source_embedding_norm
 
 
 def prepare_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
@@ -543,84 +750,45 @@ def normalize_crop_frame(crop_vision_frame : VisionFrame) -> VisionFrame:
 	model_standard_deviation = get_model_options().get('standard_deviation')
 
 	crop_vision_frame = crop_vision_frame.transpose(1, 2, 0)
-	if model_type in [ 'ghost', 'hififace', 'uniface' ]:
+
+	if model_type in [ 'ghost', 'hififace', 'hyperswap', 'uniface' ]:
 		crop_vision_frame = crop_vision_frame * model_standard_deviation + model_mean
+
 	crop_vision_frame = crop_vision_frame.clip(0, 1)
 	crop_vision_frame = crop_vision_frame[:, :, ::-1] * 255
 	return crop_vision_frame
 
 
-def get_reference_frame(source_face : Face, target_face : Face, temp_vision_frame : VisionFrame) -> VisionFrame:
-	return swap_face(source_face, target_face, temp_vision_frame)
-
-
-def process_frame(inputs : FaceSwapperInputs) -> VisionFrame:
-	reference_faces = inputs.get('reference_faces')
-	source_face = inputs.get('source_face')
-	target_vision_frame = inputs.get('target_vision_frame')
-	many_faces = sort_and_filter_faces(get_many_faces([ target_vision_frame ]))
-
-	if state_manager.get_item('face_selector_mode') == 'many':
-		if many_faces:
-			for target_face in many_faces:
-				target_vision_frame = swap_face(source_face, target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'one':
-		target_face = get_one_face(many_faces)
-		if target_face:
-			target_vision_frame = swap_face(source_face, target_face, target_vision_frame)
-	if state_manager.get_item('face_selector_mode') == 'reference':
-		similar_faces = find_similar_faces(many_faces, reference_faces, state_manager.get_item('reference_face_distance'))
-		if similar_faces:
-			for similar_face in similar_faces:
-				target_vision_frame = swap_face(source_face, similar_face, target_vision_frame)
-	return target_vision_frame
-
-
-def process_frames(source_paths : List[str], queue_payloads : List[QueuePayload], update_progress : UpdateProgress) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	source_frames = read_static_images(source_paths)
+def extract_source_face(source_vision_frames : List[VisionFrame]) -> Optional[Face]:
 	source_faces = []
 
-	for source_frame in source_frames:
-		temp_faces = get_many_faces([ source_frame ])
-		temp_faces = sort_faces_by_order(temp_faces, 'large-small')
-		if temp_faces:
-			source_faces.append(get_first(temp_faces))
-	source_face = get_average_face(source_faces)
+	if source_vision_frames:
+		for source_vision_frame in source_vision_frames:
+			temp_faces = get_static_faces([ source_vision_frame ])
+			temp_faces = sort_faces_by_order(temp_faces, 'large-small')
 
-	for queue_payload in process_manager.manage(queue_payloads):
-		target_vision_path = queue_payload['frame_path']
-		target_vision_frame = read_image(target_vision_path)
-		output_vision_frame = process_frame(
-		{
-			'reference_faces': reference_faces,
-			'source_face': source_face,
-			'target_vision_frame': target_vision_frame
-		})
-		write_image(target_vision_path, output_vision_frame)
-		update_progress(1)
+			if temp_faces:
+				source_faces.append(get_first(temp_faces))
+
+	return average_face_identity(source_faces)
 
 
-def process_image(source_paths : List[str], target_path : str, output_path : str) -> None:
-	reference_faces = get_reference_faces() if 'reference' in state_manager.get_item('face_selector_mode') else None
-	source_frames = read_static_images(source_paths)
-	source_faces = []
+def process_frame(inputs : FaceSwapperInputs) -> ProcessorOutputs:
+	reference_vision_frame = inputs.get('reference_vision_frame')
+	source_vision_frames = inputs.get('source_vision_frames')
+	target_vision_frames = inputs.get('target_vision_frames')
+	temp_vision_frame = inputs.get('temp_vision_frame')
+	temp_vision_mask = inputs.get('temp_vision_mask')
 
-	for source_frame in source_frames:
-		temp_faces = get_many_faces([ source_frame ])
-		temp_faces = sort_faces_by_order(temp_faces, 'large-small')
-		if temp_faces:
-			source_faces.append(get_first(temp_faces))
-	source_face = get_average_face(source_faces)
-	target_vision_frame = read_static_image(target_path)
-	output_vision_frame = process_frame(
-	{
-		'reference_faces': reference_faces,
-		'source_face': source_face,
-		'target_vision_frame': target_vision_frame
-	})
-	write_image(output_path, output_vision_frame)
+	target_vision_frame = get_middle(target_vision_frames)
+	source_face = extract_source_face(source_vision_frames)
+	target_faces = select_faces(reference_vision_frame, source_vision_frames, target_vision_frames)
 
+	if source_face and target_faces:
+		source_vision_frame = get_first(source_vision_frames)
 
-def process_video(source_paths : List[str], temp_frame_paths : List[str]) -> None:
-	processors.multi_process_frames(source_paths, temp_frame_paths, process_frames)
+		for target_face in target_faces:
+			target_face = scale_face(target_face, target_vision_frame, temp_vision_frame)
+			temp_vision_frame = swap_face(source_face, target_face, source_vision_frame, temp_vision_frame)
+
+	return temp_vision_frame, temp_vision_mask

facefusion/processors/modules/face_swapper/locales.py

@@ -0,0 +1,20 @@
+from facefusion.types import Locales
+
+LOCALES : Locales =\
+{
+	'en':
+	{
+		'help':
+		{
+			'model': 'choose the model responsible for swapping the face',
+			'pixel_boost': 'choose the pixel boost resolution for the face swapper',
+			'weight': 'specify the degree of weight applied to the face'
+		},
+		'uis':
+		{
+			'model_dropdown': 'FACE SWAPPER MODEL',
+			'pixel_boost_dropdown': 'FACE SWAPPER PIXEL BOOST',
+			'weight_slider': 'FACE SWAPPER WEIGHT'
+		}
+	}
+}

facefusion/processors/modules/face_swapper/types.py

@@ -0,0 +1,18 @@
+from typing import Dict, List, Literal, TypeAlias, TypedDict
+
+from facefusion.types import Mask, VisionFrame
+
+FaceSwapperInputs = TypedDict('FaceSwapperInputs',
+{
+	'reference_vision_frame' : VisionFrame,
+	'source_vision_frames' : List[VisionFrame],
+	'target_vision_frames' : List[VisionFrame],
+	'temp_vision_frame' : VisionFrame,
+	'temp_vision_mask' : Mask
+})
+
+FaceSwapperModel = Literal['blendswap_256', 'ghost_1_256', 'ghost_2_256', 'ghost_3_256', 'hififace_unofficial_256', 'hyperswap_1a_256', 'hyperswap_1b_256', 'hyperswap_1c_256', 'inswapper_128', 'inswapper_128_fp16', 'simswap_256', 'simswap_unofficial_512', 'uniface_256']
+
+FaceSwapperWeight : TypeAlias = float
+
+FaceSwapperSet : TypeAlias = Dict[FaceSwapperModel, List[str]]

facefusion/processors/modules/frame_colorizer/choices.py

@@ -0,0 +1,10 @@
+from typing import List, Sequence, get_args
+
+from facefusion.common_helper import create_int_range
+from facefusion.processors.modules.frame_colorizer.types import FrameColorizerModel
+
+frame_colorizer_models : List[FrameColorizerModel] = list(get_args(FrameColorizerModel))
+
+frame_colorizer_sizes : List[str] = [ '192x192', '256x256', '384x384', '512x512' ]
+
+frame_colorizer_blend_range : Sequence[int] = create_int_range(0, 100, 1)