Merge branch 'master' into v4

Resolve 57 conflicts keeping v4's architecture (UI removal, api app context, path isolation, streaming/codecs/rtc, workflow-mode superset, args_helper) and taking master's post-fork features so none are dropped: inference override/adjust provider hooks (+ face_swapper 3.8.1 coreml fix), is_vision_frame validity guards, resolve_temp_frame_set frame numbering with the source audio/voice trim offset, the ffmpeg color pipeline (restrict_color_transfer / convert_color_space / temp_pixel_format), ffprobe select_stream v:0 + format duration, pre_check ffprobe with dependency_not_installed, onnxruntime arena-leak guard + version lru_cache, installer cuda@12/@13 provider split, workflow_strategy disk/memory (memory default), and the dependency version bumps. content_analyser hash guard rehashed to 3c6ce25e. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01V7MdZEmd1GE8uSDTMyq34r
2026-08-07 09:38:39 +02:00 · 2026-08-05 22:12:54 +02:00
parent 11bbc89a0e 6f5e4b9d91
commit be89f03b27
50 changed files with 1032 additions and 302 deletions
@@ -79,7 +79,7 @@ def test_run_video_encode_loop(video_codec : VideoCodec, payload_type : int) ->
 	video_queue : Queue[Tuple[Time, Future[BufferPack]]] = Queue(maxsize = 30)

 	with ThreadPoolExecutor(max_workers = 1) as executor:
-		video_queue.put((0.1, executor.submit(process_video_frame, video_frame)))
+		video_queue.put((0.1, executor.submit(process_video_frame, [], video_frame)))

 		with patch('facefusion.apis.stream_video.rtc.send_video') as send_video_mock:
 			encode_loop_thread = threading.Thread(target = run_video_encode_loop, args = (rtc_peer, video_queue), daemon = True)
@@ -267,7 +267,7 @@ def test_handle_video_frame(video_codec : VideoCodec) -> None:
 	with ThreadPoolExecutor(max_workers = 1) as executor:
 		with patch('facefusion.apis.stream_video.decode_video_frame', return_value = video_frame):
 			with patch('facefusion.apis.stream_video.process_video_frame', return_value = BufferPack(buffer = video_frame.tobytes(), resolution = (426, 226))):
-				handle_video_frame(video_codec, video_decoder, video_queue, executor, bytes(), 0)
+				handle_video_frame([], video_codec, video_decoder, video_queue, executor, bytes(), 0)
 				_, video_future = video_queue.get_nowait()

 	video_buffer = video_future.result().get('buffer')
@@ -7,7 +7,7 @@ from facefusion.download import conditional_download
 from facefusion.face_creator import get_many_faces, get_one_face
 from facefusion.face_store import clear_faces
 from facefusion.face_tracker import create_face_tracks, select_face_track, track_faces
-from facefusion.vision import read_static_video_chunk, read_static_video_frame
+from facefusion.vision import read_static_video_frame, select_video_frames
 from .assert_helper import get_test_example_file, get_test_examples_directory


@@ -47,8 +47,7 @@ def before_each() -> None:

 def test_track_faces() -> None:
 	target_path = get_test_example_file('target-240p.mp4')
-	video_frame_chunk = read_static_video_chunk(target_path, 0, 7)
-	target_vision_frames = [ video_frame_chunk.get(frame_number) for frame_number in sorted(video_frame_chunk) ]
+	target_vision_frames = select_video_frames(target_path, 3, 3)
 	empty_vision_frame = numpy.zeros_like(get_first(target_vision_frames))

 	target_vision_frames[2] = empty_vision_frame
@@ -58,7 +57,7 @@ def test_track_faces() -> None:

 	assert len(track_faces(target_vision_frames, 0.3)) == 1

-	target_vision_frames = [ video_frame_chunk.get(frame_number) for frame_number in sorted(video_frame_chunk)[:5] ]
+	target_vision_frames = select_video_frames(target_path, 3, 3)[:5]
 	target_vision_frames[0] = empty_vision_frame
 	target_vision_frames[1] = empty_vision_frame
 	target_vision_frames[2] = empty_vision_frame
@@ -7,7 +7,7 @@ import facefusion.ffmpeg
 from facefusion import ffmpeg, ffmpeg_builder, process_manager, state_manager
 from facefusion.download import conditional_download
 from facefusion.ffmpeg import concat_video, extract_frames, merge_video, read_audio_buffer, replace_audio, restore_audio, sanitize_audio, sanitize_image, sanitize_video, spawn_frames
-from facefusion.ffprobe import probe_entries
+from facefusion.ffprobe import probe_audio_entries, probe_video_entries
 from facefusion.filesystem import copy_file, is_image
 from facefusion.temp_helper import clear_temp_directory, create_temp_directory, get_temp_file_path, resolve_temp_frame_paths
 from facefusion.types import EncoderSet
@@ -253,10 +253,10 @@ def test_sanitize_audio() -> None:
 	]

 	assert sanitize_audio(file_content, output_paths[0], 'strict') is True
-	assert probe_entries(output_paths[0], [ 'codec_name' ]).get('codec_name') == 'mp3'
+	assert probe_audio_entries(output_paths[0], [ 'codec_name' ]).get('codec_name') == 'mp3'

 	assert sanitize_audio(file_content, output_paths[1], 'moderate') is True
-	assert probe_entries(output_paths[1], [ 'codec_name' ]).get('codec_name') == 'pcm_s16le'
+	assert probe_audio_entries(output_paths[1], [ 'codec_name' ]).get('codec_name') == 'pcm_s16le'


 def test_sanitize_image() -> None:
@@ -278,9 +278,9 @@ def test_sanitize_video() -> None:
 	]

 	assert sanitize_video(file_content, output_paths[0], 'strict') is True
-	assert probe_entries(output_paths[0], [ 'codec_name' ]).get('codec_name') == 'h264'
+	assert probe_video_entries(output_paths[0], [ 'codec_name' ]).get('codec_name') == 'h264'

 	assert sanitize_video(file_content, output_paths[1], 'moderate') is True
-	assert probe_entries(output_paths[1], [ 'codec_name' ]).get('codec_name') == 'hevc'
+	assert probe_video_entries(output_paths[1], [ 'codec_name' ]).get('codec_name') == 'hevc'


@@ -26,13 +26,14 @@ def before_all() -> None:
 			ffmpeg_builder.set_output(get_test_example_file('source-48000khz-2ch.wav'))
 		)
 	)
-	ffmpeg.run_ffmpeg(
-		ffmpeg_builder.chain(
-			ffmpeg_builder.set_input(get_test_example_file('target-240p.mp4')),
-			ffmpeg_builder.set_video_duration(1),
-			ffmpeg_builder.set_output(get_test_example_file('target-240p-1s.mov'))
+	for video_format in [ 'mkv', 'mov' ]:
+		ffmpeg.run_ffmpeg(
+			ffmpeg_builder.chain(
+				ffmpeg_builder.set_input(get_test_example_file('target-240p.mp4')),
+				ffmpeg_builder.set_video_duration(1),
+				ffmpeg_builder.set_output(get_test_example_file('target-240p-1s.' + video_format))
+			)
 		)
-	)


 def test_extract_audio_metadata() -> None:
@@ -40,7 +41,7 @@ def test_extract_audio_metadata() -> None:

 	assert audio_metadata.get('sample_rate') == 44100
 	assert audio_metadata.get('channel_total') == 1
-	assert audio_metadata.get('frame_total') == 167039
+	assert audio_metadata.get('frame_total') == 167040
 	assert audio_metadata.get('bit_rate') == 128000

 	audio_metadata = extract_audio_metadata(get_test_example_file('source-48000khz-2ch.wav'))
@@ -48,7 +49,7 @@ def test_extract_audio_metadata() -> None:
 	assert audio_metadata.get('sample_rate') == 48000
 	assert audio_metadata.get('channel_total') == 2
 	assert audio_metadata.get('frame_total') == 91200
-	assert audio_metadata.get('bit_rate') == 1536000
+	assert audio_metadata.get('bit_rate') == 1536328


 def test_extract_video_metadata() -> None:
@@ -57,7 +58,15 @@ def test_extract_video_metadata() -> None:
 	assert video_metadata.get('fps') == 25.0
 	assert video_metadata.get('duration') == 10.8
 	assert video_metadata.get('resolution') == (426, 226)
-	assert video_metadata.get('bit_rate') == 138754
+	assert video_metadata.get('bit_rate') == 141981
+	assert video_metadata.get('color_transfer') == 'smpte170m'
+
+	video_metadata = extract_video_metadata(get_test_example_file('target-240p-1s.mkv'))
+
+	assert video_metadata.get('fps') == 25.0
+	assert video_metadata.get('duration') == 1.0
+	assert video_metadata.get('frame_total') == 25
+	assert video_metadata.get('resolution') == (426, 226)

 	video_metadata = extract_video_metadata(get_test_example_file('target-240p-1s.mov'))

@@ -1,7 +1,7 @@
 from shutil import which

 from facefusion import ffprobe_builder
-from facefusion.ffprobe_builder import chain, format_to_key_value, run, set_input, show_entries
+from facefusion.ffprobe_builder import chain, format_to_key_value, run, select_stream, set_input, show_stream_entries


 def test_run() -> None:
@@ -10,15 +10,21 @@ def test_run() -> None:

 def test_chain() -> None:
 	assert chain(
-		ffprobe_builder.show_entries([ 'sample_rate' ]),
+		ffprobe_builder.select_stream('a:0'),
+		ffprobe_builder.show_stream_entries([ 'sample_rate' ]),
 		ffprobe_builder.format_to_key_value(),
 		ffprobe_builder.set_input('audio.mp3')
-	) == [ '-show_entries', 'stream=sample_rate', '-of', 'default=noprint_wrappers=1', '-i', 'audio.mp3' ]
+	) == [ '-select_streams', 'a:0', '-show_entries', 'stream=sample_rate', '-of', 'default=noprint_wrappers=1', '-i', 'audio.mp3' ]


-def test_show_entries() -> None:
-	assert show_entries([ 'duration' ]) == [ '-show_entries', 'stream=duration' ]
-	assert show_entries([ 'duration', 'sample_rate']) == [ '-show_entries', 'stream=duration,sample_rate' ]
+def test_select_stream() -> None:
+	assert select_stream('a:0') == [ '-select_streams', 'a:0' ]
+	assert select_stream('v:0') == [ '-select_streams', 'v:0' ]
+
+
+def test_show_stream_entries() -> None:
+	assert show_stream_entries([ 'duration' ]) == [ '-show_entries', 'stream=duration' ]
+	assert show_stream_entries([ 'duration', 'sample_rate' ]) == [ '-show_entries', 'stream=duration,sample_rate' ]


 def test_format_to_key_value() -> None:
@@ -0,0 +1,73 @@
+import pytest
+
+from facefusion import process_manager
+from facefusion.download import conditional_download
+from facefusion.frame_store import clear_frames, get_frame_store, reduce_frames, select_frame_set, set_frame
+from facefusion.vision import read_video_frame
+from .assert_helper import get_test_example_file, get_test_examples_directory
+
+
+@pytest.fixture(scope = 'module', autouse = True)
+def before_all() -> None:
+	process_manager.start()
+	conditional_download(get_test_examples_directory(),
+	[
+		'https://github.com/facefusion/facefusion-assets/releases/download/examples-3.0.0/target-240p.mp4'
+	])
+
+
+@pytest.fixture(scope = 'function', autouse = True)
+def before_each() -> None:
+	clear_frames('reader-1')
+	clear_frames('reader-2')
+
+
+def test_get_frame_store() -> None:
+	frame_store = get_frame_store('reader-1')
+
+	assert frame_store == {}
+	assert get_frame_store('reader-1') is frame_store
+
+
+def test_set_frame() -> None:
+	target_frame = read_video_frame(get_test_example_file('target-240p.mp4'), 0)
+
+	set_frame('reader-1', 5, target_frame)
+
+	assert get_frame_store('reader-1').get(5) is target_frame
+
+
+def test_select_frame_set() -> None:
+	first_frame = read_video_frame(get_test_example_file('target-240p.mp4'), 0)
+	fifth_frame = read_video_frame(get_test_example_file('target-240p.mp4'), 5)
+
+	set_frame('reader-1', 2, first_frame)
+	set_frame('reader-1', 5, fifth_frame)
+
+	assert sorted(select_frame_set('reader-1', 0, 4)) == [ 2 ]
+	assert select_frame_set('reader-1', 0, 4).get(2) is first_frame
+	assert sorted(select_frame_set('reader-1', 2, 6)) == [ 2, 5 ]
+	assert select_frame_set('reader-1', 2, 6).get(5) is fifth_frame
+	assert select_frame_set('reader-1', 8, 12) == {}
+
+
+def test_reduce_frames() -> None:
+	target_frame = read_video_frame(get_test_example_file('target-240p.mp4'), 0)
+
+	for frame_number in range(0, 10):
+		set_frame('reader-1', frame_number, target_frame)
+
+	reduce_frames('reader-1', 4, 6)
+
+	assert sorted(get_frame_store('reader-1')) == [ 4, 5, 6 ]
+
+
+def test_clear_frames() -> None:
+	target_frame = read_video_frame(get_test_example_file('target-240p.mp4'), 0)
+
+	set_frame('reader-1', 0, target_frame)
+	set_frame('reader-2', 0, target_frame)
+	clear_frames('reader-1')
+
+	assert get_frame_store('reader-1') == {}
+	assert get_frame_store('reader-2').get(0) is target_frame
@@ -0,0 +1,203 @@
+import tempfile
+
+import numpy
+import pytest
+
+from facefusion import ffmpeg, ffmpeg_builder, process_manager, state_manager
+from facefusion.common_helper import is_linux, is_macos, is_windows
+from facefusion.download import conditional_download
+from facefusion.ffprobe import extract_video_metadata
+from facefusion.frame_store import get_frame_store
+from facefusion.temp_helper import create_temp_directory, get_temp_file_path
+from facefusion.video_manager import clear_video_pool, close_video_reader, close_video_writer, collect_video_frames, conditional_seek_video_reader, drain_video_reader, get_reader, get_writer, read_video_frame, read_video_frames, seek_video_reader, write_video_frame
+from .assert_helper import get_test_example_file, get_test_examples_directory
+
+
+@pytest.fixture(scope = 'module', autouse = True)
+def before_all() -> None:
+	process_manager.start()
+	conditional_download(get_test_examples_directory(),
+	[
+		'https://github.com/facefusion/facefusion-assets/releases/download/examples-3.0.0/target-240p.mp4'
+	])
+
+	for video_fps in [ 25, 30 ]:
+		ffmpeg.run_ffmpeg(
+			ffmpeg_builder.chain(
+				ffmpeg_builder.set_input(get_test_example_file('target-240p.mp4')),
+				ffmpeg_builder.set_video_fps(video_fps),
+				ffmpeg_builder.set_output(get_test_example_file('target-240p-' + str(video_fps) + 'fps.mp4'))
+			)
+		)
+
+	state_manager.init_item('temp_path', tempfile.gettempdir())
+	state_manager.init_item('temp_frame_format', 'png')
+	state_manager.init_item('temp_pixel_format', 'bgr24')
+	state_manager.init_item('output_video_encoder', 'libx264')
+	state_manager.init_item('output_video_quality', 80)
+	state_manager.init_item('output_video_preset', 'veryfast')
+
+
+@pytest.fixture(scope = 'function', autouse = True)
+def before_each() -> None:
+	clear_video_pool()
+
+
+def test_get_reader() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+	video_metadata = video_reader.get('metadata')
+
+	assert video_metadata.get('resolution') == (426, 226)
+	assert video_metadata.get('fps') == 25.0
+	assert video_metadata.get('frame_total') == 270
+
+	assert get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame') is video_reader
+	assert not get_reader(get_test_example_file('target-240p-25fps.mp4'), 'select_video_frames').get('id') == video_reader.get('id')
+
+
+def test_conditional_seek_video_reader() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+	video_frames = {}
+
+	for frame_number in range(30):
+		video_frames[frame_number] = read_video_frame(video_reader)
+
+	for frame_number in [ 5, 17, 29 ]:
+		conditional_seek_video_reader(video_reader, frame_number)
+
+		assert numpy.array_equal(read_video_frame(video_reader), video_frames.get(frame_number)) is True
+
+
+def test_seek_video_reader() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+	video_frames = {}
+
+	for frame_number in range(30):
+		video_frames[frame_number] = read_video_frame(video_reader)
+
+	for frame_number in [ 5, 17, 29 ]:
+		seek_video_reader(video_reader, frame_number)
+
+		assert numpy.array_equal(read_video_frame(video_reader), video_frames.get(frame_number)) is True
+
+
+def test_drain_video_reader() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+
+	drain_video_reader(video_reader, 10)
+
+	assert video_reader.get('frame_number') == 10
+
+	vision_frame = read_video_frame(video_reader)
+	seek_video_reader(video_reader, 10)
+
+	assert numpy.array_equal(vision_frame, read_video_frame(video_reader)) is True
+
+
+def test_read_video_frame() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+
+	assert read_video_frame(video_reader).shape == (226, 426, 3)
+	assert video_reader.get('frame_number') == 1
+
+	seek_video_reader(video_reader, 269)
+
+	assert read_video_frame(video_reader).shape == (226, 426, 3)
+	assert read_video_frame(video_reader) is None
+
+
+def test_read_video_frames() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'read_video_frame')
+
+	assert sorted(read_video_frames(video_reader, 0, 4)) == [ 0, 1, 2, 3, 4 ]
+
+	frame_number = video_reader.get('frame_number')
+
+	assert sorted(read_video_frames(video_reader, 1, 3)) == [ 1, 2, 3 ]
+	assert video_reader.get('frame_number') == frame_number
+
+	read_video_frames(video_reader, 21, 25)
+
+	assert min(get_frame_store(video_reader.get('id'))) == 17
+	assert max(get_frame_store(video_reader.get('id'))) == 25
+	assert sorted(read_video_frames(video_reader, 268, 275)) == [ 268, 269 ]
+
+
+def test_collect_video_frames() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'select_video_frames')
+
+	collect_video_frames(video_reader, 20, 24)
+
+	assert sorted(get_frame_store(video_reader.get('id'))) == [ 20, 21, 22, 23, 24 ]
+	assert video_reader.get('frame_number') == 25
+
+
+def test_close_video_reader() -> None:
+	video_reader = get_reader(get_test_example_file('target-240p-25fps.mp4'), 'select_video_frames')
+
+	read_video_frames(video_reader, 0, 4)
+	close_video_reader(video_reader)
+
+	if is_windows():
+		assert video_reader.get('process').returncode == 1
+
+	if is_linux() or is_macos():
+		assert video_reader.get('process').returncode == -9
+
+
+def test_get_writer() -> None:
+	target_path = get_test_example_file('target-240p-25fps.mp4')
+	create_temp_directory(state_manager.get_temp_path(), target_path)
+	video_writer = get_writer(target_path, 25.0, (426, 226), (426, 226), 25.0)
+
+	assert get_writer(target_path, 25.0, (426, 226), (426, 226), 25.0) is video_writer
+
+
+def test_write_video_frame() -> None:
+	target_path = get_test_example_file('target-240p-25fps.mp4')
+	create_temp_directory(state_manager.get_temp_path(), target_path)
+	video_reader = get_reader(target_path, 'read_video_frame')
+	video_writer = get_writer(target_path, 25.0, (426, 226), (426, 226), 25.0)
+
+	for frame_number in range(25):
+		write_video_frame(video_writer, read_video_frame(video_reader))
+
+	assert close_video_writer(video_writer) is True
+
+	video_metadata = extract_video_metadata(get_temp_file_path(state_manager.get_temp_path(), target_path))
+
+	assert video_metadata.get('duration') == 1.0
+	assert video_metadata.get('frame_total') == 25
+	assert video_metadata.get('fps') == 25.0
+	assert video_metadata.get('resolution') == (426, 226)
+	assert video_metadata.get('color_transfer') == 'bt709'
+
+
+def test_close_video_writer() -> None:
+	target_path = get_test_example_file('target-240p-30fps.mp4')
+	create_temp_directory(state_manager.get_temp_path(), target_path)
+	video_reader = get_reader(target_path, 'read_video_frame')
+	video_writer = get_writer(target_path, 30.0, (426, 226), (426, 226), 30.0)
+
+	write_video_frame(video_writer, read_video_frame(video_reader))
+
+	assert close_video_writer(video_writer) is True
+
+
+def test_clear_video_pool() -> None:
+	target_path = get_test_example_file('target-240p-25fps.mp4')
+	create_temp_directory(state_manager.get_temp_path(), target_path)
+	video_reader = get_reader(target_path, 'select_video_frames')
+	video_writer = get_writer(target_path, 25.0, (426, 226), (426, 226), 25.0)
+
+	read_video_frames(video_reader, 0, 4)
+	write_video_frame(video_writer, read_video_frame(video_reader))
+	clear_video_pool()
+
+	if is_windows():
+		assert video_reader.get('process').returncode == 1
+
+	if is_linux() or is_macos():
+		assert video_reader.get('process').returncode == -9
+
+	assert video_writer.get('process').returncode == 0
@@ -4,7 +4,7 @@ import pytest

 from facefusion import ffmpeg, ffmpeg_builder, process_manager
 from facefusion.download import conditional_download
-from facefusion.vision import calculate_histogram_difference, count_video_frame_total, detect_image_resolution, detect_video_duration, detect_video_fps, detect_video_resolution, match_frame_color, normalize_resolution, pack_resolution, predict_video_frame_total, read_image, read_video_chunk, read_video_frame, restrict_image_resolution, restrict_trim_video_frame, restrict_video_fps, restrict_video_resolution, scale_resolution, select_video_frames, unpack_resolution, write_image
+from facefusion.vision import calculate_histogram_difference, count_video_frame_total, detect_image_resolution, detect_video_duration, detect_video_fps, detect_video_resolution, match_frame_color, normalize_resolution, pack_resolution, predict_video_frame_total, read_image, read_video_frame, restrict_image_resolution, restrict_trim_video_frame, restrict_video_fps, restrict_video_resolution, scale_resolution, select_video_frames, unpack_resolution, write_image
 from .assert_helper import get_test_example_file, get_test_examples_directory, get_test_output_path, prepare_test_output_directory


@@ -136,11 +136,6 @@ def test_read_video_frame() -> None:
 	assert read_video_frame('invalid') is None


-def test_read_video_chunk() -> None:
-	assert len(read_video_chunk(get_test_example_file('target-240p-25fps.mp4'), 1, 40)) == 40
-	assert read_video_chunk('invalid', 1, 40) == {}
-
-
 def test_select_video_frames() -> None:
 	assert len(select_video_frames(get_test_example_file('target-240p-25fps.mp4'), 50, 5)) == 11
 	assert len(select_video_frames(get_test_example_file('target-240p-25fps.mp4'), 1, 5)) == 11