use ideal pipesize for our task

* fix index bug

* fix rounding bug

facefusion/face_store.py

@@ -11,7 +11,7 @@ FACE_STORE : FaceStore = {}
 
 def get_faces(vision_frame : VisionFrame) -> Optional[List[Face]]:
 	if numpy.any(vision_frame):
-		vision_hash = create_hash(vision_frame.tobytes())
+		vision_hash = create_hash(vision_frame.data)
 
 		if FACE_STORE.get(vision_hash):
 			return FACE_STORE.get(vision_hash).get('faces')
@@ -21,7 +21,7 @@ def get_faces(vision_frame : VisionFrame) -> Optional[List[Face]]:
 
 def set_faces(vision_frame : VisionFrame, faces : List[Face]) -> None:
 	if numpy.any(vision_frame):
-		vision_hash = create_hash(vision_frame.tobytes())
+		vision_hash = create_hash(vision_frame.data)
 		FACE_STORE.setdefault(vision_hash,
 		{
 			'lock': threading.Lock()
@@ -30,7 +30,7 @@ def set_faces(vision_frame : VisionFrame, faces : List[Face]) -> None:
 
 def resolve_lock(vision_frame : VisionFrame) -> threading.Lock:
 	if numpy.any(vision_frame):
-		vision_hash = create_hash(vision_frame.tobytes())
+		vision_hash = create_hash(vision_frame.data)
 		return FACE_STORE.setdefault(vision_hash,
 		{
 			'lock': threading.Lock()

facefusion/ffmpeg.py

@@ -66,7 +66,7 @@ def run_ffmpeg(commands : List[Command]) -> 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)
+	return subprocess.Popen(commands, stdin = subprocess.PIPE, stdout = subprocess.PIPE, pipesize = 1024 * 1024)
 
 
 def create_video_reader(video_path : str, frame_number : int, video_metadata : VideoReaderMetadata) -> subprocess.Popen[bytes]:

facefusion/ffprobe.py

@@ -74,7 +74,7 @@ def extract_audio_metadata(audio_path : str) -> AudioMetadata:
 
 	duration = float(format_entries.get('duration'))
 	sample_rate = int(audio_entries.get('sample_rate'))
-	frame_total = int(duration * sample_rate)
+	frame_total = round(duration * sample_rate)
 	channel_total = int(audio_entries.get('channels'))
 	bit_rate = int(format_entries.get('bit_rate'))
 
@@ -101,7 +101,7 @@ def extract_video_metadata(video_path : str) -> VideoMetadata:
 
 	duration = float(format_entries.get('duration'))
 	fps = extract_video_fps(video_entries.get('r_frame_rate'))
-	frame_total = int(duration * fps)
+	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'))

facefusion/uis/components/preview_options.py

@@ -26,8 +26,9 @@ def render() -> None:
 		'visible': False
 	}
 	if is_video(state_manager.get_item('target_path')):
+		video_frame_total = count_video_frame_total(state_manager.get_item('target_path'))
 		preview_frame_slider_options['value'] = state_manager.get_item('reference_frame_number')
-		preview_frame_slider_options['maximum'] = count_video_frame_total(state_manager.get_item('target_path'))
+		preview_frame_slider_options['maximum'] = video_frame_total - 1
 		preview_frame_slider_options['visible'] = True
 	PREVIEW_FRAME_SLIDER = gradio.Slider(**preview_frame_slider_options)
 	with gradio.Row():
@@ -57,5 +58,5 @@ def listen() -> None:
 def update_preview_frame_slider() -> gradio.Slider:
 	if is_video(state_manager.get_item('target_path')):
 		video_frame_total = count_video_frame_total(state_manager.get_item('target_path'))
-		return gradio.Slider(maximum = video_frame_total, visible = True)
+		return gradio.Slider(maximum = video_frame_total - 1, visible = True)
 	return gradio.Slider(value = 0, visible = False)

facefusion/uis/components/webcam.py

@@ -106,7 +106,7 @@ def start(webcam_device_id : int, webcam_mode : WebcamMode, webcam_resolution :
 				yield capture_vision_frame
 			if webcam_mode in [ 'udp', 'v4l2' ]:
 				try:
-					stream.stdin.write(capture_vision_frame.tobytes())
+					stream.stdin.write(capture_vision_frame.data)
 				except Exception:
 					pass
 

facefusion/video_manager.py

@@ -35,7 +35,8 @@ def get_reader(video_path : str, context : str) -> VideoReader:
 
 
 def conditional_seek_video_reader(video_reader : VideoReader, frame_number : int = 0) -> None:
-	frame_number = min(video_reader.get('metadata').get('frame_total'), frame_number)
+	frame_total = video_reader.get('metadata').get('frame_total')
+	frame_number = min(frame_total - 1, frame_number)
 	skip_total = frame_number - video_reader.get('frame_number')
 	skip_margin = 128
 
@@ -132,7 +133,7 @@ def get_writer(video_path : str, temp_video_fps : Fps, temp_video_resolution : R
 
 
 def write_video_frame(video_writer : VideoWriter, vision_frame : VisionFrame) -> None:
-	video_writer.get('process').stdin.write(vision_frame.tobytes())
+	video_writer.get('process').stdin.write(vision_frame.data)
 
 
 def close_video_writer(video_writer : VideoWriter) -> bool:

tests/test_ffprobe.py

@@ -41,7 +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'))