Skip to content

Industry leading face manipulation platform

Introduce video manager to handle broken videos

henryruhs62483fdfad5b
diff --git a/facefusion/face_store.py b/facefusion/face_store.py
index 518b9e3c..59e6d5d1 100644
--- a/facefusion/face_store.py
+++ b/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()
diff --git a/facefusion/processors/modules/age_modifier.py b/facefusion/processors/modules/age_modifier.py
index 925808b6..47197442 100755
--- a/facefusion/processors/modules/age_modifier.py
+++ b/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':
diff --git a/facefusion/processors/modules/deep_swapper.py b/facefusion/processors/modules/deep_swapper.py
index 14e23708..54438c71 100755
--- a/facefusion/processors/modules/deep_swapper.py
+++ b/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':
diff --git a/facefusion/processors/modules/expression_restorer.py b/facefusion/processors/modules/expression_restorer.py
index 4d301a40..0e307ddc 100755
--- a/facefusion/processors/modules/expression_restorer.py
+++ b/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':
diff --git a/facefusion/processors/modules/face_debugger.py b/facefusion/processors/modules/face_debugger.py
index e4397f23..ea53b7d5 100755
--- a/facefusion/processors/modules/face_debugger.py
+++ b/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()
diff --git a/facefusion/processors/modules/face_editor.py b/facefusion/processors/modules/face_editor.py
index 3f494d8d..2414d146 100755
--- a/facefusion/processors/modules/face_editor.py
+++ b/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':
diff --git a/facefusion/processors/modules/face_enhancer.py b/facefusion/processors/modules/face_enhancer.py
index 1787d692..ad1b0a52 100755
--- a/facefusion/processors/modules/face_enhancer.py
+++ b/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':
diff --git a/facefusion/processors/modules/face_swapper.py b/facefusion/processors/modules/face_swapper.py
index 7ce0ed42..1df1fa02 100755
--- a/facefusion/processors/modules/face_swapper.py
+++ b/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()
diff --git a/facefusion/processors/modules/frame_colorizer.py b/facefusion/processors/modules/frame_colorizer.py
index 12a1942c..abf3fc91 100644
--- a/facefusion/processors/modules/frame_colorizer.py
+++ b/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':
diff --git a/facefusion/processors/modules/frame_enhancer.py b/facefusion/processors/modules/frame_enhancer.py
index 5b653a75..27a9b6c1 100644
--- a/facefusion/processors/modules/frame_enhancer.py
+++ b/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':
diff --git a/facefusion/processors/modules/lip_syncer.py b/facefusion/processors/modules/lip_syncer.py
index 1b6faa2e..ac38cbb9 100755
--- a/facefusion/processors/modules/lip_syncer.py
+++ b/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':
diff --git a/facefusion/types.py b/facefusion/types.py
index def3aa09..57210b01 100755
--- a/facefusion/types.py
+++ b/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]
diff --git a/facefusion/video_manager.py b/facefusion/video_manager.py
new file mode 100644
index 00000000..f38bc19a
--- /dev/null
+++ b/facefusion/video_manager.py
@@ -0,0 +1,21 @@
+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:
+ global VIDEO_POOL_SET
+
+ for video_capture in VIDEO_POOL_SET.values():
+ video_capture.release()
+
+ VIDEO_POOL_SET.clear()
diff --git a/facefusion/vision.py b/facefusion/vision.py
index d03e3834..b146170c 100644
--- a/facefusion/vision.py
+++ b/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()
- video_capture.release()
+
+ with thread_semaphore():
+ 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))
- video_capture.release()
- return video_frame_total
+ with thread_semaphore():
+ video_frame_total = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
+ 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)
- video_capture.release()
- return video_fps
+ with thread_semaphore():
+ video_fps = video_capture.get(cv2.CAP_PROP_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)
- height = video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
- video_capture.release()
- return int(width), int(height)
+ with thread_semaphore():
+ width = video_capture.get(cv2.CAP_PROP_FRAME_WIDTH)
+ height = video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
+ return int(width), int(height)
+
return None