Merge pull request #886 from facefusion/feat/video_manager

feat/video manager

facefusion/core.py

@@ -6,7 +6,7 @@ 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 cli_helper, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, process_manager, state_manager, video_manager, voice_extractor, wording
 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
@@ -468,8 +468,10 @@ def process_video(start_time : float) -> ErrorCode:
 		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')):
+				video_manager.clear_video_pool()
 				logger.debug(wording.get('replacing_audio_succeed'), __name__)
 			else:
+				video_manager.clear_video_pool()
 				if is_process_stopping():
 					process_manager.end()
 					return 4
@@ -477,8 +479,10 @@ def process_video(start_time : float) -> ErrorCode:
 				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):
+				video_manager.clear_video_pool()
 				logger.debug(wording.get('restoring_audio_succeed'), __name__)
 			else:
+				video_manager.clear_video_pool()
 				if is_process_stopping():
 					process_manager.end()
 					return 4

facefusion/face_store.py

@@ -30,7 +30,7 @@ def set_static_faces(vision_frame : VisionFrame, faces : List[Face]) -> None:
 
 
 def clear_static_faces() -> None:
-	FACE_STORE['static_faces'] = {}
+	FACE_STORE['static_faces'].clear()
 
 
 def create_frame_hash(vision_frame : VisionFrame) -> Optional[str]:
@@ -53,4 +53,4 @@ def append_reference_face(name : str, face : Face) -> None:
 
 
 def clear_reference_faces() -> None:
-	FACE_STORE['reference_faces'] = {}
+	FACE_STORE['reference_faces'].clear()

facefusion/processors/modules/age_modifier.py

@@ -9,7 +9,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_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
@@ -115,6 +115,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/deep_swapper.py

@@ -9,7 +9,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_manager, wording
 from facefusion.common_helper import create_int_metavar
 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
@@ -311,6 +311,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/expression_restorer.py

@@ -8,7 +8,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_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
@@ -129,6 +129,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/face_debugger.py

@@ -7,7 +7,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, logger, process_manager, state_manager, video_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
@@ -56,6 +56,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.cache_clear()
+	video_manager.clear_video_pool()
 	if state_manager.get_item('video_memory_strategy') == 'strict':
 		content_analyser.clear_inference_pool()
 		face_classifier.clear_inference_pool()

facefusion/processors/modules/face_editor.py

@@ -8,7 +8,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_manager, wording
 from facefusion.common_helper import create_float_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
@@ -182,6 +182,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/face_enhancer.py

@@ -8,7 +8,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_manager, wording
 from facefusion.common_helper import create_float_metavar, 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
@@ -275,6 +275,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/face_swapper.py

@@ -8,7 +8,7 @@ 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 import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_manager, wording
 from facefusion.common_helper import get_first
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
 from facefusion.execution import has_execution_provider
@@ -406,9 +406,10 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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()

facefusion/processors/modules/frame_colorizer.py

@@ -8,7 +8,7 @@ import numpy
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
 import facefusion.processors.core as processors
-from facefusion import config, content_analyser, inference_manager, logger, process_manager, state_manager, wording
+from facefusion import config, content_analyser, inference_manager, logger, process_manager, state_manager, video_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
@@ -188,6 +188,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/frame_enhancer.py

@@ -8,7 +8,7 @@ import numpy
 import facefusion.jobs.job_manager
 import facefusion.jobs.job_store
 import facefusion.processors.core as processors
-from facefusion import config, content_analyser, inference_manager, logger, process_manager, state_manager, wording
+from facefusion import config, content_analyser, inference_manager, logger, process_manager, state_manager, video_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
@@ -450,6 +450,7 @@ def pre_process(mode : ProcessMode) -> bool:
 
 def post_process() -> None:
 	read_static_image.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':

facefusion/processors/modules/lip_syncer.py

@@ -8,7 +8,7 @@ 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, voice_extractor, wording
+from facefusion import config, content_analyser, face_classifier, face_detector, face_landmarker, face_masker, face_recognizer, inference_manager, logger, process_manager, state_manager, video_manager, voice_extractor, wording
 from facefusion.audio import create_empty_audio_frame, get_voice_frame, read_static_voice
 from facefusion.common_helper import get_first
 from facefusion.download import conditional_download_hashes, conditional_download_sources, resolve_download_url
@@ -127,6 +127,7 @@ def pre_process(mode : ProcessMode) -> bool:
 def post_process() -> None:
 	read_static_image.cache_clear()
 	read_static_voice.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':

facefusion/types.py

@@ -1,6 +1,7 @@
 from collections import namedtuple
 from typing import Any, Callable, Dict, List, Literal, Optional, Tuple, TypeAlias, TypedDict
 
+import cv2
 import numpy
 from numpy.typing import NDArray
 from onnxruntime import InferenceSession
@@ -49,6 +50,7 @@ FaceStore = TypedDict('FaceStore',
 	'static_faces' : FaceSet,
 	'reference_faces' : FaceSet
 })
+VideoPoolSet : TypeAlias = Dict[str, cv2.VideoCapture]
 
 VisionFrame : TypeAlias = NDArray[Any]
 Mask : TypeAlias = NDArray[Any]

facefusion/video_manager.py

@@ -0,0 +1,19 @@
+import cv2
+
+from facefusion.types import VideoPoolSet
+
+VIDEO_POOL_SET : VideoPoolSet = {}
+
+
+def get_video_capture(video_path : str) -> cv2.VideoCapture:
+	if video_path not in VIDEO_POOL_SET:
+		VIDEO_POOL_SET[video_path] = cv2.VideoCapture(video_path)
+
+	return VIDEO_POOL_SET.get(video_path)
+
+
+def clear_video_pool() -> None:
+	for video_capture in VIDEO_POOL_SET.values():
+		video_capture.release()
+
+	VIDEO_POOL_SET.clear()

facefusion/vision.py

@@ -9,7 +9,9 @@ from cv2.typing import Size
 import facefusion.choices
 from facefusion.common_helper import is_windows
 from facefusion.filesystem import get_file_extension, is_image, is_video
+from facefusion.thread_helper import thread_semaphore
 from facefusion.types import Duration, Fps, Orientation, Resolution, VisionFrame
+from facefusion.video_manager import get_video_capture
 
 
 @lru_cache()
@@ -81,24 +83,30 @@ def create_image_resolutions(resolution : Resolution) -> List[str]:
 
 def read_video_frame(video_path : str, frame_number : int = 0) -> Optional[VisionFrame]:
 	if is_video(video_path):
-		video_capture = cv2.VideoCapture(video_path)
+		video_capture = get_video_capture(video_path)
+
 		if video_capture.isOpened():
 			frame_total = video_capture.get(cv2.CAP_PROP_FRAME_COUNT)
-			video_capture.set(cv2.CAP_PROP_POS_FRAMES, min(frame_total, frame_number - 1))
+
-			has_vision_frame, vision_frame = video_capture.read()
+			with thread_semaphore():
-			video_capture.release()
+				video_capture.set(cv2.CAP_PROP_POS_FRAMES, min(frame_total, frame_number - 1))
+				has_vision_frame, vision_frame = video_capture.read()
+
 			if has_vision_frame:
 				return vision_frame
+
 	return None
 
 
 def count_video_frame_total(video_path : str) -> int:
 	if is_video(video_path):
-		video_capture = cv2.VideoCapture(video_path)
+		video_capture = get_video_capture(video_path)
+
 		if video_capture.isOpened():
-			video_frame_total = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
+			with thread_semaphore():
-			video_capture.release()
+				video_frame_total = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
-			return video_frame_total
+				return video_frame_total
+
 	return 0
 
 
@@ -112,11 +120,13 @@ def predict_video_frame_total(video_path : str, fps : Fps, trim_frame_start : in
 
 def detect_video_fps(video_path : str) -> Optional[float]:
 	if is_video(video_path):
-		video_capture = cv2.VideoCapture(video_path)
+		video_capture = get_video_capture(video_path)
+
 		if video_capture.isOpened():
-			video_fps = video_capture.get(cv2.CAP_PROP_FPS)
+			with thread_semaphore():
-			video_capture.release()
+				video_fps = video_capture.get(cv2.CAP_PROP_FPS)
-			return video_fps
+				return video_fps
+
 	return None
 
 
@@ -163,12 +173,14 @@ def restrict_trim_frame(video_path : str, trim_frame_start : Optional[int], trim
 
 def detect_video_resolution(video_path : str) -> Optional[Resolution]:
 	if is_video(video_path):
-		video_capture = cv2.VideoCapture(video_path)
+		video_capture = get_video_capture(video_path)
+
 		if video_capture.isOpened():
-			width = video_capture.get(cv2.CAP_PROP_FRAME_WIDTH)
+			with thread_semaphore():
-			height = video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
+				width = video_capture.get(cv2.CAP_PROP_FRAME_WIDTH)
-			video_capture.release()
+				height = video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
-			return int(width), int(height)
+				return int(width), int(height)
+
 	return None