vettorazi commited on
Commit
86a87f2
1 Parent(s): 67b8ef2

Upload 34 files

Browse files
models/inswapper_128.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4a3f08c753cb72d04e10aa0f7dbe3deebbf39567d4ead6dce08e98aa49e16af
3
+ size 554253681
roop/Scripts/python.exe ADDED
Binary file (269 kB). View file
 
roop/Scripts/pythonw.exe ADDED
Binary file (257 kB). View file
 
roop/__init__.py ADDED
File without changes
roop/__pycache__/__init__.cpython-310.pyc ADDED
Binary file (140 Bytes). View file
 
roop/__pycache__/capturer.cpython-310.pyc ADDED
Binary file (814 Bytes). View file
 
roop/__pycache__/core.cpython-310.pyc ADDED
Binary file (9.09 kB). View file
 
roop/__pycache__/face_analyser.cpython-310.pyc ADDED
Binary file (1.14 kB). View file
 
roop/__pycache__/globals.cpython-310.pyc ADDED
Binary file (536 Bytes). View file
 
roop/__pycache__/metadata.cpython-310.pyc ADDED
Binary file (175 Bytes). View file
 
roop/__pycache__/predicter.cpython-310.pyc ADDED
Binary file (1.22 kB). View file
 
roop/__pycache__/typing.cpython-310.pyc ADDED
Binary file (278 Bytes). View file
 
roop/__pycache__/ui.cpython-310.pyc ADDED
Binary file (8.4 kB). View file
 
roop/__pycache__/utilities.cpython-310.pyc ADDED
Binary file (5.59 kB). View file
 
roop/capturer.py ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import Any
2
+ import cv2
3
+
4
+
5
+ def get_video_frame(video_path: str, frame_number: int = 0) -> Any:
6
+ capture = cv2.VideoCapture(video_path)
7
+ frame_total = capture.get(cv2.CAP_PROP_FRAME_COUNT)
8
+ capture.set(cv2.CAP_PROP_POS_FRAMES, min(frame_total, frame_number - 1))
9
+ has_frame, frame = capture.read()
10
+ capture.release()
11
+ if has_frame:
12
+ return frame
13
+ return None
14
+
15
+
16
+ def get_video_frame_total(video_path: str) -> int:
17
+ capture = cv2.VideoCapture(video_path)
18
+ video_frame_total = int(capture.get(cv2.CAP_PROP_FRAME_COUNT))
19
+ capture.release()
20
+ return video_frame_total
roop/core.py ADDED
@@ -0,0 +1,238 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+
3
+ import os
4
+ import sys
5
+ # single thread doubles cuda performance - needs to be set before torch import
6
+ if any(arg.startswith('--execution-provider') for arg in sys.argv):
7
+ os.environ['OMP_NUM_THREADS'] = '1'
8
+ # reduce tensorflow log level
9
+ os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
10
+ import warnings
11
+ from typing import List
12
+ import platform
13
+ import signal
14
+ import shutil
15
+ import argparse
16
+ import torch
17
+ import onnxruntime
18
+ import tensorflow
19
+
20
+ import roop.globals
21
+ import roop.metadata
22
+ import roop.ui as ui
23
+ from roop.predicter import predict_image, predict_video
24
+ from roop.processors.frame.core import get_frame_processors_modules
25
+ from roop.utilities import has_image_extension, is_image, is_video, detect_fps, create_video, extract_frames, get_temp_frame_paths, restore_audio, create_temp, move_temp, clean_temp, normalize_output_path
26
+
27
+ if 'ROCMExecutionProvider' in roop.globals.execution_providers:
28
+ del torch
29
+
30
+ warnings.filterwarnings('ignore', category=FutureWarning, module='insightface')
31
+ warnings.filterwarnings('ignore', category=UserWarning, module='torchvision')
32
+
33
+
34
+ def parse_args() -> None:
35
+ signal.signal(signal.SIGINT, lambda signal_number, frame: destroy())
36
+ program = argparse.ArgumentParser()
37
+ program.add_argument('-s', '--source', help='select an source image', dest='source_path')
38
+ program.add_argument('-t', '--target', help='select an target image or video', dest='target_path')
39
+ program.add_argument('-o', '--output', help='select output file or directory', dest='output_path')
40
+ program.add_argument('--frame-processor', help='pipeline of frame processors', dest='frame_processor', default=['face_swapper'], choices=['face_swapper', 'face_enhancer'], nargs='+')
41
+ program.add_argument('--keep-fps', help='keep original fps', dest='keep_fps', action='store_true', default=False)
42
+ program.add_argument('--keep-audio', help='keep original audio', dest='keep_audio', action='store_true', default=True)
43
+ program.add_argument('--keep-frames', help='keep temporary frames', dest='keep_frames', action='store_true', default=False)
44
+ program.add_argument('--many-faces', help='process every face', dest='many_faces', action='store_true', default=False)
45
+ program.add_argument('--video-encoder', help='adjust output video encoder', dest='video_encoder', default='libx264', choices=['libx264', 'libx265', 'libvpx-vp9'])
46
+ program.add_argument('--video-quality', help='adjust output video quality', dest='video_quality', type=int, default=18, choices=range(52), metavar='[0-51]')
47
+ program.add_argument('--max-memory', help='maximum amount of RAM in GB', dest='max_memory', type=int, default=suggest_max_memory())
48
+ program.add_argument('--execution-provider', help='execution provider', dest='execution_provider', default=['cpu'], choices=suggest_execution_providers(), nargs='+')
49
+ program.add_argument('--execution-threads', help='number of execution threads', dest='execution_threads', type=int, default=suggest_execution_threads())
50
+ program.add_argument('-v', '--version', action='version', version=f'{roop.metadata.name} {roop.metadata.version}')
51
+
52
+ # register deprecated args
53
+ program.add_argument('-f', '--face', help=argparse.SUPPRESS, dest='source_path_deprecated')
54
+ program.add_argument('--cpu-cores', help=argparse.SUPPRESS, dest='cpu_cores_deprecated', type=int)
55
+ program.add_argument('--gpu-vendor', help=argparse.SUPPRESS, dest='gpu_vendor_deprecated')
56
+ program.add_argument('--gpu-threads', help=argparse.SUPPRESS, dest='gpu_threads_deprecated', type=int)
57
+
58
+ args = program.parse_args()
59
+
60
+ roop.globals.source_path = args.source_path
61
+ roop.globals.target_path = args.target_path
62
+ roop.globals.output_path = normalize_output_path(roop.globals.source_path, roop.globals.target_path, args.output_path)
63
+ roop.globals.frame_processors = args.frame_processor
64
+ roop.globals.headless = args.source_path or args.target_path or args.output_path
65
+ roop.globals.keep_fps = args.keep_fps
66
+ roop.globals.keep_audio = args.keep_audio
67
+ roop.globals.keep_frames = args.keep_frames
68
+ roop.globals.many_faces = args.many_faces
69
+ roop.globals.video_encoder = args.video_encoder
70
+ roop.globals.video_quality = args.video_quality
71
+ roop.globals.max_memory = args.max_memory
72
+ roop.globals.execution_providers = decode_execution_providers(args.execution_provider)
73
+ roop.globals.execution_threads = args.execution_threads
74
+
75
+ # translate deprecated args
76
+ if args.source_path_deprecated:
77
+ print('\033[33mArgument -f and --face are deprecated. Use -s and --source instead.\033[0m')
78
+ roop.globals.source_path = args.source_path_deprecated
79
+ roop.globals.output_path = normalize_output_path(args.source_path_deprecated, roop.globals.target_path, args.output_path)
80
+ if args.cpu_cores_deprecated:
81
+ print('\033[33mArgument --cpu-cores is deprecated. Use --execution-threads instead.\033[0m')
82
+ roop.globals.execution_threads = args.cpu_cores_deprecated
83
+ if args.gpu_vendor_deprecated == 'apple':
84
+ print('\033[33mArgument --gpu-vendor apple is deprecated. Use --execution-provider coreml instead.\033[0m')
85
+ roop.globals.execution_providers = decode_execution_providers(['coreml'])
86
+ if args.gpu_vendor_deprecated == 'nvidia':
87
+ print('\033[33mArgument --gpu-vendor nvidia is deprecated. Use --execution-provider cuda instead.\033[0m')
88
+ roop.globals.execution_providers = decode_execution_providers(['cuda'])
89
+ if args.gpu_vendor_deprecated == 'amd':
90
+ print('\033[33mArgument --gpu-vendor amd is deprecated. Use --execution-provider cuda instead.\033[0m')
91
+ roop.globals.execution_providers = decode_execution_providers(['rocm'])
92
+ if args.gpu_threads_deprecated:
93
+ print('\033[33mArgument --gpu-threads is deprecated. Use --execution-threads instead.\033[0m')
94
+ roop.globals.execution_threads = args.gpu_threads_deprecated
95
+
96
+
97
+ def encode_execution_providers(execution_providers: List[str]) -> List[str]:
98
+ return [execution_provider.replace('ExecutionProvider', '').lower() for execution_provider in execution_providers]
99
+
100
+
101
+ def decode_execution_providers(execution_providers: List[str]) -> List[str]:
102
+ return [provider for provider, encoded_execution_provider in zip(onnxruntime.get_available_providers(), encode_execution_providers(onnxruntime.get_available_providers()))
103
+ if any(execution_provider in encoded_execution_provider for execution_provider in execution_providers)]
104
+
105
+
106
+ def suggest_max_memory() -> int:
107
+ if platform.system().lower() == 'darwin':
108
+ return 4
109
+ return 16
110
+
111
+
112
+ def suggest_execution_providers() -> List[str]:
113
+ return encode_execution_providers(onnxruntime.get_available_providers())
114
+
115
+
116
+ def suggest_execution_threads() -> int:
117
+ if 'DmlExecutionProvider' in roop.globals.execution_providers:
118
+ return 1
119
+ if 'ROCMExecutionProvider' in roop.globals.execution_providers:
120
+ return 1
121
+ return 8
122
+
123
+
124
+ def limit_resources() -> None:
125
+ # prevent tensorflow memory leak
126
+ gpus = tensorflow.config.experimental.list_physical_devices('GPU')
127
+ for gpu in gpus:
128
+ tensorflow.config.experimental.set_memory_growth(gpu, True)
129
+ # limit memory usage
130
+ if roop.globals.max_memory:
131
+ memory = roop.globals.max_memory * 1024 ** 3
132
+ if platform.system().lower() == 'darwin':
133
+ memory = roop.globals.max_memory * 1024 ** 6
134
+ if platform.system().lower() == 'windows':
135
+ import ctypes
136
+ kernel32 = ctypes.windll.kernel32
137
+ kernel32.SetProcessWorkingSetSize(-1, ctypes.c_size_t(memory), ctypes.c_size_t(memory))
138
+ else:
139
+ import resource
140
+ resource.setrlimit(resource.RLIMIT_DATA, (memory, memory))
141
+
142
+
143
+ def release_resources() -> None:
144
+ if 'CUDAExecutionProvider' in roop.globals.execution_providers:
145
+ torch.cuda.empty_cache()
146
+
147
+
148
+ def pre_check() -> bool:
149
+ if sys.version_info < (3, 9):
150
+ update_status('Python version is not supported - please upgrade to 3.9 or higher.')
151
+ return False
152
+ if not shutil.which('ffmpeg'):
153
+ update_status('ffmpeg is not installed.')
154
+ return False
155
+ return True
156
+
157
+
158
+ def update_status(message: str, scope: str = 'ROOP.CORE') -> None:
159
+ print(f'[{scope}] {message}')
160
+ if not roop.globals.headless:
161
+ ui.update_status(message)
162
+
163
+
164
+ def start() -> None:
165
+ for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
166
+ if not frame_processor.pre_start():
167
+ return
168
+ # process image to image
169
+ if has_image_extension(roop.globals.target_path):
170
+ if predict_image(roop.globals.target_path):
171
+ destroy()
172
+ shutil.copy2(roop.globals.target_path, roop.globals.output_path)
173
+ for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
174
+ update_status('Progressing...', frame_processor.NAME)
175
+ frame_processor.process_image(roop.globals.source_path, roop.globals.output_path, roop.globals.output_path)
176
+ release_resources()
177
+ if is_image(roop.globals.target_path):
178
+ update_status('Processing to image succeed!')
179
+ else:
180
+ update_status('Processing to image failed!')
181
+ return
182
+ # process image to videos
183
+ if predict_video(roop.globals.target_path):
184
+ destroy()
185
+ update_status('Creating temp resources...')
186
+ create_temp(roop.globals.target_path)
187
+ update_status('Extracting frames...')
188
+ extract_frames(roop.globals.target_path)
189
+ temp_frame_paths = get_temp_frame_paths(roop.globals.target_path)
190
+ for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
191
+ update_status('Progressing...', frame_processor.NAME)
192
+ frame_processor.process_video(roop.globals.source_path, temp_frame_paths)
193
+ release_resources()
194
+ # handles fps
195
+ if roop.globals.keep_fps:
196
+ update_status('Detecting fps...')
197
+ fps = detect_fps(roop.globals.target_path)
198
+ update_status(f'Creating video with {fps} fps...')
199
+ create_video(roop.globals.target_path, fps)
200
+ else:
201
+ update_status('Creating video with 30.0 fps...')
202
+ create_video(roop.globals.target_path)
203
+ # handle audio
204
+ if roop.globals.keep_audio:
205
+ if roop.globals.keep_fps:
206
+ update_status('Restoring audio...')
207
+ else:
208
+ update_status('Restoring audio might cause issues as fps are not kept...')
209
+ restore_audio(roop.globals.target_path, roop.globals.output_path)
210
+ else:
211
+ move_temp(roop.globals.target_path, roop.globals.output_path)
212
+ # clean and validate
213
+ clean_temp(roop.globals.target_path)
214
+ if is_video(roop.globals.target_path):
215
+ update_status('Processing to video succeed!')
216
+ else:
217
+ update_status('Processing to video failed!')
218
+
219
+
220
+ def destroy() -> None:
221
+ if roop.globals.target_path:
222
+ clean_temp(roop.globals.target_path)
223
+ quit()
224
+
225
+
226
+ def run() -> None:
227
+ parse_args()
228
+ if not pre_check():
229
+ return
230
+ for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
231
+ if not frame_processor.pre_check():
232
+ return
233
+ limit_resources()
234
+ if roop.globals.headless:
235
+ start()
236
+ else:
237
+ window = ui.init(start, destroy)
238
+ window.mainloop()
roop/face_analyser.py ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import Any
2
+ import insightface
3
+
4
+ import roop.globals
5
+ from roop.typing import Frame
6
+
7
+ FACE_ANALYSER = None
8
+
9
+
10
+ def get_face_analyser() -> Any:
11
+ global FACE_ANALYSER
12
+
13
+ if FACE_ANALYSER is None:
14
+ FACE_ANALYSER = insightface.app.FaceAnalysis(name='buffalo_l', providers=roop.globals.execution_providers)
15
+ FACE_ANALYSER.prepare(ctx_id=0, det_size=(640, 640))
16
+ return FACE_ANALYSER
17
+
18
+
19
+ def get_one_face(frame: Frame) -> Any:
20
+ face = get_face_analyser().get(frame)
21
+ try:
22
+ return min(face, key=lambda x: x.bbox[0])
23
+ except ValueError:
24
+ return None
25
+
26
+
27
+ def get_many_faces(frame: Frame) -> Any:
28
+ try:
29
+ return get_face_analyser().get(frame)
30
+ except IndexError:
31
+ return None
roop/globals.py ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+
3
+ source_path = None
4
+ target_path = None
5
+ output_path = None
6
+ frame_processors: List[str] = []
7
+ keep_fps = None
8
+ keep_audio = None
9
+ keep_frames = None
10
+ many_faces = None
11
+ video_encoder = None
12
+ video_quality = None
13
+ max_memory = None
14
+ execution_providers: List[str] = []
15
+ execution_threads = None
16
+ headless = None
17
+ log_level = 'error'
roop/metadata.py ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ name = 'roop'
2
+ version = '1.0.1'
roop/predicter.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import numpy
2
+ import opennsfw2
3
+ from PIL import Image
4
+
5
+ from roop.typing import Frame
6
+
7
+ MAX_PROBABILITY = 0.85
8
+
9
+
10
+ def predict_frame(target_frame: Frame) -> bool:
11
+ image = Image.fromarray(target_frame)
12
+ image = opennsfw2.preprocess_image(image, opennsfw2.Preprocessing.YAHOO)
13
+ model = opennsfw2.make_open_nsfw_model()
14
+ views = numpy.expand_dims(image, axis=0)
15
+ _, probability = model.predict(views)[0]
16
+ return probability > MAX_PROBABILITY
17
+
18
+
19
+ def predict_image(target_path: str) -> bool:
20
+ return opennsfw2.predict_image(target_path) > MAX_PROBABILITY
21
+
22
+
23
+ def predict_video(target_path: str) -> bool:
24
+ _, probabilities = opennsfw2.predict_video_frames(video_path=target_path, frame_interval=100)
25
+ return any(probability > MAX_PROBABILITY for probability in probabilities)
roop/processors/__init__.py ADDED
File without changes
roop/processors/__pycache__/__init__.cpython-310.pyc ADDED
Binary file (151 Bytes). View file
 
roop/processors/frame/__init__.py ADDED
File without changes
roop/processors/frame/__pycache__/__init__.cpython-310.pyc ADDED
Binary file (157 Bytes). View file
 
roop/processors/frame/__pycache__/core.cpython-310.pyc ADDED
Binary file (2.21 kB). View file
 
roop/processors/frame/__pycache__/face_swapper.cpython-310.pyc ADDED
Binary file (3.11 kB). View file
 
roop/processors/frame/core.py ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import sys
2
+ import importlib
3
+ from concurrent.futures import ThreadPoolExecutor
4
+ from types import ModuleType
5
+ from typing import Any, List, Callable
6
+ from tqdm import tqdm
7
+
8
+ import roop
9
+
10
+ FRAME_PROCESSORS_MODULES: List[ModuleType] = []
11
+ FRAME_PROCESSORS_INTERFACE = [
12
+ 'pre_check',
13
+ 'pre_start',
14
+ 'process_frame',
15
+ 'process_image',
16
+ 'process_video'
17
+ ]
18
+
19
+
20
+ def load_frame_processor_module(frame_processor: str) -> Any:
21
+ try:
22
+ frame_processor_module = importlib.import_module(f'roop.processors.frame.{frame_processor}')
23
+ for method_name in FRAME_PROCESSORS_INTERFACE:
24
+ if not hasattr(frame_processor_module, method_name):
25
+ sys.exit()
26
+ except ImportError:
27
+ sys.exit()
28
+ return frame_processor_module
29
+
30
+
31
+ def get_frame_processors_modules(frame_processors: List[str]) -> List[ModuleType]:
32
+ global FRAME_PROCESSORS_MODULES
33
+
34
+ if not FRAME_PROCESSORS_MODULES:
35
+ for frame_processor in frame_processors:
36
+ frame_processor_module = load_frame_processor_module(frame_processor)
37
+ FRAME_PROCESSORS_MODULES.append(frame_processor_module)
38
+ return FRAME_PROCESSORS_MODULES
39
+
40
+
41
+ def multi_process_frame(source_path: str, temp_frame_paths: List[str], process_frames: Callable[[str, List[str], Any], None], progress: Any = None) -> None:
42
+ with ThreadPoolExecutor(max_workers=roop.globals.execution_threads) as executor:
43
+ futures = []
44
+ for path in temp_frame_paths:
45
+ future = executor.submit(process_frames, source_path, [path], progress)
46
+ futures.append(future)
47
+ for future in futures:
48
+ future.result()
49
+
50
+
51
+ def process_video(source_path: str, frame_paths: list[str], process_frames: Callable[[str, List[str], Any], None]) -> None:
52
+ progress_bar_format = '{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}, {rate_fmt}{postfix}]'
53
+ total = len(frame_paths)
54
+ with tqdm(total=total, desc='Processing', unit='frame', dynamic_ncols=True, bar_format=progress_bar_format) as progress:
55
+ progress.set_postfix({'execution_providers': roop.globals.execution_providers, 'threads': roop.globals.execution_threads, 'memory': roop.globals.max_memory})
56
+ multi_process_frame(source_path, frame_paths, process_frames, progress)
roop/processors/frame/face_enhancer.py ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import Any, List
2
+ import cv2
3
+ import threading
4
+ import gfpgan
5
+
6
+ import roop.globals
7
+ import roop.processors.frame.core
8
+ from roop.core import update_status
9
+ from roop.face_analyser import get_one_face
10
+ from roop.typing import Frame, Face
11
+ from roop.utilities import conditional_download, resolve_relative_path, is_image, is_video
12
+
13
+ FACE_ENHANCER = None
14
+ THREAD_SEMAPHORE = threading.Semaphore()
15
+ THREAD_LOCK = threading.Lock()
16
+ NAME = 'ROOP.FACE-ENHANCER'
17
+
18
+
19
+ def pre_check() -> bool:
20
+ download_directory_path = resolve_relative_path('../models')
21
+ conditional_download(download_directory_path, ['https://huggingface.co/henryruhs/roop/resolve/main/GFPGANv1.4.pth'])
22
+ return True
23
+
24
+
25
+ def pre_start() -> bool:
26
+ if not is_image(roop.globals.target_path) and not is_video(roop.globals.target_path):
27
+ update_status('Select an image or video for target path.', NAME)
28
+ return False
29
+ return True
30
+
31
+
32
+ def get_face_enhancer() -> Any:
33
+ global FACE_ENHANCER
34
+
35
+ with THREAD_LOCK:
36
+ if FACE_ENHANCER is None:
37
+ model_path = resolve_relative_path('../models/GFPGANv1.4.pth')
38
+ # todo: set models path https://github.com/TencentARC/GFPGAN/issues/399
39
+ FACE_ENHANCER = gfpgan.GFPGANer(model_path=model_path, upscale=1) # type: ignore[attr-defined]
40
+ return FACE_ENHANCER
41
+
42
+
43
+ def enhance_face(temp_frame: Frame) -> Frame:
44
+ with THREAD_SEMAPHORE:
45
+ _, _, temp_frame = get_face_enhancer().enhance(
46
+ temp_frame,
47
+ paste_back=True
48
+ )
49
+ return temp_frame
50
+
51
+
52
+ def process_frame(source_face: Face, temp_frame: Frame) -> Frame:
53
+ target_face = get_one_face(temp_frame)
54
+ if target_face:
55
+ temp_frame = enhance_face(temp_frame)
56
+ return temp_frame
57
+
58
+
59
+ def process_frames(source_path: str, temp_frame_paths: List[str], progress: Any = None) -> None:
60
+ for temp_frame_path in temp_frame_paths:
61
+ temp_frame = cv2.imread(temp_frame_path)
62
+ result = process_frame(None, temp_frame)
63
+ cv2.imwrite(temp_frame_path, result)
64
+ if progress:
65
+ progress.update(1)
66
+
67
+
68
+ def process_image(source_path: str, target_path: str, output_path: str) -> None:
69
+ target_frame = cv2.imread(target_path)
70
+ result = process_frame(None, target_frame)
71
+ cv2.imwrite(output_path, result)
72
+
73
+
74
+ def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
75
+ roop.processors.frame.core.process_video(None, temp_frame_paths, process_frames)
roop/processors/frame/face_swapper.py ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import Any, List
2
+ import cv2
3
+ import insightface
4
+ import threading
5
+
6
+ import roop.globals
7
+ import roop.processors.frame.core
8
+ from roop.core import update_status
9
+ from roop.face_analyser import get_one_face, get_many_faces
10
+ from roop.typing import Face, Frame
11
+ from roop.utilities import conditional_download, resolve_relative_path, is_image, is_video
12
+
13
+ FACE_SWAPPER = None
14
+ THREAD_LOCK = threading.Lock()
15
+ NAME = 'ROOP.FACE-SWAPPER'
16
+
17
+
18
+ def pre_check() -> bool:
19
+ download_directory_path = resolve_relative_path('../models')
20
+ conditional_download(download_directory_path, ['https://huggingface.co/henryruhs/roop/resolve/main/inswapper_128.onnx'])
21
+ return True
22
+
23
+
24
+ def pre_start() -> bool:
25
+ if not is_image(roop.globals.source_path):
26
+ update_status('Select an image for source path.', NAME)
27
+ return False
28
+ elif not get_one_face(cv2.imread(roop.globals.source_path)):
29
+ update_status('No face in source path detected.', NAME)
30
+ return False
31
+ if not is_image(roop.globals.target_path) and not is_video(roop.globals.target_path):
32
+ update_status('Select an image or video for target path.', NAME)
33
+ return False
34
+ return True
35
+
36
+
37
+ def get_face_swapper() -> Any:
38
+ global FACE_SWAPPER
39
+
40
+ with THREAD_LOCK:
41
+ if FACE_SWAPPER is None:
42
+ model_path = resolve_relative_path('../models/inswapper_128.onnx')
43
+ FACE_SWAPPER = insightface.model_zoo.get_model(model_path, providers=roop.globals.execution_providers)
44
+ return FACE_SWAPPER
45
+
46
+
47
+ def swap_face(source_face: Face, target_face: Face, temp_frame: Frame) -> Frame:
48
+ return get_face_swapper().get(temp_frame, target_face, source_face, paste_back=True)
49
+
50
+
51
+ def process_frame(source_face: Face, temp_frame: Frame) -> Frame:
52
+ if roop.globals.many_faces:
53
+ many_faces = get_many_faces(temp_frame)
54
+ if many_faces:
55
+ for target_face in many_faces:
56
+ temp_frame = swap_face(source_face, target_face, temp_frame)
57
+ else:
58
+ target_face = get_one_face(temp_frame)
59
+ if target_face:
60
+ temp_frame = swap_face(source_face, target_face, temp_frame)
61
+ return temp_frame
62
+
63
+
64
+ def process_frames(source_path: str, temp_frame_paths: List[str], progress: Any = None) -> None:
65
+ source_face = get_one_face(cv2.imread(source_path))
66
+ for temp_frame_path in temp_frame_paths:
67
+ temp_frame = cv2.imread(temp_frame_path)
68
+ try:
69
+ result = process_frame(source_face, temp_frame)
70
+ cv2.imwrite(temp_frame_path, result)
71
+ except Exception as exception:
72
+ print(exception)
73
+ pass
74
+ if progress:
75
+ progress.update(1)
76
+
77
+
78
+ def process_image(source_path: str, target_path: str, output_path: str) -> None:
79
+ source_face = get_one_face(cv2.imread(source_path))
80
+ target_frame = cv2.imread(target_path)
81
+ result = process_frame(source_face, target_frame)
82
+ cv2.imwrite(output_path, result)
83
+
84
+
85
+ def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
86
+ roop.processors.frame.core.process_video(source_path, temp_frame_paths, process_frames)
roop/pyvenv.cfg ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ home = C:\Users\danie\AppData\Local\Microsoft\WindowsApps\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0
2
+ include-system-site-packages = false
3
+ version = 3.10.11
roop/typing.py ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ from typing import Any
2
+
3
+ from insightface.app.common import Face
4
+ import numpy
5
+
6
+ Face = Face
7
+ Frame = numpy.ndarray[Any, Any]
roop/ui.json ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "CTk": {
3
+ "fg_color": ["gray95", "gray10"]
4
+ },
5
+ "CTkToplevel": {
6
+ "fg_color": ["gray95", "gray10"]
7
+ },
8
+ "CTkFrame": {
9
+ "corner_radius": 6,
10
+ "border_width": 0,
11
+ "fg_color": ["gray90", "gray13"],
12
+ "top_fg_color": ["gray85", "gray16"],
13
+ "border_color": ["gray65", "gray28"]
14
+ },
15
+ "CTkButton": {
16
+ "corner_radius": 6,
17
+ "border_width": 0,
18
+ "fg_color": ["#3a7ebf", "#1f538d"],
19
+ "hover_color": ["#325882", "#14375e"],
20
+ "border_color": ["#3E454A", "#949A9F"],
21
+ "text_color": ["#DCE4EE", "#DCE4EE"],
22
+ "text_color_disabled": ["gray74", "gray60"]
23
+ },
24
+ "CTkLabel": {
25
+ "corner_radius": 0,
26
+ "fg_color": "transparent",
27
+ "text_color": ["gray14", "gray84"]
28
+ },
29
+ "CTkEntry": {
30
+ "corner_radius": 6,
31
+ "border_width": 2,
32
+ "fg_color": ["#F9F9FA", "#343638"],
33
+ "border_color": ["#979DA2", "#565B5E"],
34
+ "text_color": ["gray14", "gray84"],
35
+ "placeholder_text_color": ["gray52", "gray62"]
36
+ },
37
+ "CTkCheckbox": {
38
+ "corner_radius": 6,
39
+ "border_width": 3,
40
+ "fg_color": ["#3a7ebf", "#1f538d"],
41
+ "border_color": ["#3E454A", "#949A9F"],
42
+ "hover_color": ["#325882", "#14375e"],
43
+ "checkmark_color": ["#DCE4EE", "gray90"],
44
+ "text_color": ["gray14", "gray84"],
45
+ "text_color_disabled": ["gray60", "gray45"]
46
+ },
47
+ "CTkSwitch": {
48
+ "corner_radius": 1000,
49
+ "border_width": 3,
50
+ "button_length": 0,
51
+ "fg_color": ["#939BA2", "#4A4D50"],
52
+ "progress_color": ["#3a7ebf", "#1f538d"],
53
+ "button_color": ["gray36", "#D5D9DE"],
54
+ "button_hover_color": ["gray20", "gray100"],
55
+ "text_color": ["gray14", "gray84"],
56
+ "text_color_disabled": ["gray60", "gray45"]
57
+ },
58
+ "CTkRadiobutton": {
59
+ "corner_radius": 1000,
60
+ "border_width_checked": 6,
61
+ "border_width_unchecked": 3,
62
+ "fg_color": ["#3a7ebf", "#1f538d"],
63
+ "border_color": ["#3E454A", "#949A9F"],
64
+ "hover_color": ["#325882", "#14375e"],
65
+ "text_color": ["gray14", "gray84"],
66
+ "text_color_disabled": ["gray60", "gray45"]
67
+ },
68
+ "CTkProgressBar": {
69
+ "corner_radius": 1000,
70
+ "border_width": 0,
71
+ "fg_color": ["#939BA2", "#4A4D50"],
72
+ "progress_color": ["#3a7ebf", "#1f538d"],
73
+ "border_color": ["gray", "gray"]
74
+ },
75
+ "CTkSlider": {
76
+ "corner_radius": 1000,
77
+ "button_corner_radius": 1000,
78
+ "border_width": 6,
79
+ "button_length": 0,
80
+ "fg_color": ["#939BA2", "#4A4D50"],
81
+ "progress_color": ["gray40", "#AAB0B5"],
82
+ "button_color": ["#3a7ebf", "#1f538d"],
83
+ "button_hover_color": ["#325882", "#14375e"]
84
+ },
85
+ "CTkOptionMenu": {
86
+ "corner_radius": 6,
87
+ "fg_color": ["#3a7ebf", "#1f538d"],
88
+ "button_color": ["#325882", "#14375e"],
89
+ "button_hover_color": ["#234567", "#1e2c40"],
90
+ "text_color": ["#DCE4EE", "#DCE4EE"],
91
+ "text_color_disabled": ["gray74", "gray60"]
92
+ },
93
+ "CTkComboBox": {
94
+ "corner_radius": 6,
95
+ "border_width": 2,
96
+ "fg_color": ["#F9F9FA", "#343638"],
97
+ "border_color": ["#979DA2", "#565B5E"],
98
+ "button_color": ["#979DA2", "#565B5E"],
99
+ "button_hover_color": ["#6E7174", "#7A848D"],
100
+ "text_color": ["gray14", "gray84"],
101
+ "text_color_disabled": ["gray50", "gray45"]
102
+ },
103
+ "CTkScrollbar": {
104
+ "corner_radius": 1000,
105
+ "border_spacing": 4,
106
+ "fg_color": "transparent",
107
+ "button_color": ["gray55", "gray41"],
108
+ "button_hover_color": ["gray40", "gray53"]
109
+ },
110
+ "CTkSegmentedButton": {
111
+ "corner_radius": 6,
112
+ "border_width": 2,
113
+ "fg_color": ["#979DA2", "gray29"],
114
+ "selected_color": ["#3a7ebf", "#1f538d"],
115
+ "selected_hover_color": ["#325882", "#14375e"],
116
+ "unselected_color": ["#979DA2", "gray29"],
117
+ "unselected_hover_color": ["gray70", "gray41"],
118
+ "text_color": ["#DCE4EE", "#DCE4EE"],
119
+ "text_color_disabled": ["gray74", "gray60"]
120
+ },
121
+ "CTkTextbox": {
122
+ "corner_radius": 6,
123
+ "border_width": 0,
124
+ "fg_color": ["gray100", "gray20"],
125
+ "border_color": ["#979DA2", "#565B5E"],
126
+ "text_color": ["gray14", "gray84"],
127
+ "scrollbar_button_color": ["gray55", "gray41"],
128
+ "scrollbar_button_hover_color": ["gray40", "gray53"]
129
+ },
130
+ "CTkScrollableFrame": {
131
+ "label_fg_color": ["gray80", "gray21"]
132
+ },
133
+ "DropdownMenu": {
134
+ "fg_color": ["gray90", "gray20"],
135
+ "hover_color": ["gray75", "gray28"],
136
+ "text_color": ["gray14", "gray84"]
137
+ },
138
+ "CTkFont": {
139
+ "macOS": {
140
+ "family": "Avenir",
141
+ "size": 12,
142
+ "weight": "normal"
143
+ },
144
+ "Windows": {
145
+ "family": "Corbel",
146
+ "size": 12,
147
+ "weight": "normal"
148
+ },
149
+ "Linux": {
150
+ "family": "Montserrat",
151
+ "size": 12,
152
+ "weight": "normal"
153
+ }
154
+ },
155
+ "RoopDonate": {
156
+ "text_color": ["gray74", "gray60"]
157
+ }
158
+ }
roop/ui.py ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ import webbrowser
3
+ import customtkinter as ctk
4
+ from typing import Callable, Tuple
5
+ import cv2
6
+ from PIL import Image, ImageOps
7
+
8
+ import roop.globals
9
+ import roop.metadata
10
+ from roop.face_analyser import get_one_face
11
+ from roop.capturer import get_video_frame, get_video_frame_total
12
+ from roop.predicter import predict_frame
13
+ from roop.processors.frame.core import get_frame_processors_modules
14
+ from roop.utilities import is_image, is_video, resolve_relative_path
15
+
16
+ ROOT = None
17
+ ROOT_HEIGHT = 700
18
+ ROOT_WIDTH = 600
19
+
20
+ PREVIEW = None
21
+ PREVIEW_MAX_HEIGHT = 700
22
+ PREVIEW_MAX_WIDTH = 1200
23
+
24
+ RECENT_DIRECTORY_SOURCE = None
25
+ RECENT_DIRECTORY_TARGET = None
26
+ RECENT_DIRECTORY_OUTPUT = None
27
+
28
+ preview_label = None
29
+ preview_slider = None
30
+ source_label = None
31
+ target_label = None
32
+ status_label = None
33
+
34
+
35
+ def init(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.CTk:
36
+ global ROOT, PREVIEW
37
+
38
+ ROOT = create_root(start, destroy)
39
+ PREVIEW = create_preview(ROOT)
40
+
41
+ return ROOT
42
+
43
+
44
+ def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.CTk:
45
+ global source_label, target_label, status_label
46
+
47
+ ctk.deactivate_automatic_dpi_awareness()
48
+ ctk.set_appearance_mode('system')
49
+ ctk.set_default_color_theme(resolve_relative_path('ui.json'))
50
+
51
+ root = ctk.CTk()
52
+ root.minsize(ROOT_WIDTH, ROOT_HEIGHT)
53
+ root.title(f'{roop.metadata.name} {roop.metadata.version}')
54
+ root.configure()
55
+ root.protocol('WM_DELETE_WINDOW', lambda: destroy())
56
+
57
+ source_label = ctk.CTkLabel(root, text=None)
58
+ source_label.place(relx=0.1, rely=0.1, relwidth=0.3, relheight=0.25)
59
+
60
+ target_label = ctk.CTkLabel(root, text=None)
61
+ target_label.place(relx=0.6, rely=0.1, relwidth=0.3, relheight=0.25)
62
+
63
+ source_button = ctk.CTkButton(root, text='Select a face', cursor='hand2', command=lambda: select_source_path())
64
+ source_button.place(relx=0.1, rely=0.4, relwidth=0.3, relheight=0.1)
65
+
66
+ target_button = ctk.CTkButton(root, text='Select a target', cursor='hand2', command=lambda: select_target_path())
67
+ target_button.place(relx=0.6, rely=0.4, relwidth=0.3, relheight=0.1)
68
+
69
+ keep_fps_value = ctk.BooleanVar(value=roop.globals.keep_fps)
70
+ keep_fps_checkbox = ctk.CTkSwitch(root, text='Keep fps', variable=keep_fps_value, cursor='hand2', command=lambda: setattr(roop.globals, 'keep_fps', not roop.globals.keep_fps))
71
+ keep_fps_checkbox.place(relx=0.1, rely=0.6)
72
+
73
+ keep_frames_value = ctk.BooleanVar(value=roop.globals.keep_frames)
74
+ keep_frames_switch = ctk.CTkSwitch(root, text='Keep frames', variable=keep_frames_value, cursor='hand2', command=lambda: setattr(roop.globals, 'keep_frames', keep_frames_value.get()))
75
+ keep_frames_switch.place(relx=0.1, rely=0.65)
76
+
77
+ keep_audio_value = ctk.BooleanVar(value=roop.globals.keep_audio)
78
+ keep_audio_switch = ctk.CTkSwitch(root, text='Keep audio', variable=keep_audio_value, cursor='hand2', command=lambda: setattr(roop.globals, 'keep_audio', keep_audio_value.get()))
79
+ keep_audio_switch.place(relx=0.6, rely=0.6)
80
+
81
+ many_faces_value = ctk.BooleanVar(value=roop.globals.many_faces)
82
+ many_faces_switch = ctk.CTkSwitch(root, text='Many faces', variable=many_faces_value, cursor='hand2', command=lambda: setattr(roop.globals, 'many_faces', many_faces_value.get()))
83
+ many_faces_switch.place(relx=0.6, rely=0.65)
84
+
85
+ start_button = ctk.CTkButton(root, text='Start', cursor='hand2', command=lambda: select_output_path(start))
86
+ start_button.place(relx=0.15, rely=0.75, relwidth=0.2, relheight=0.05)
87
+
88
+ stop_button = ctk.CTkButton(root, text='Destroy', cursor='hand2', command=lambda: destroy())
89
+ stop_button.place(relx=0.4, rely=0.75, relwidth=0.2, relheight=0.05)
90
+
91
+ preview_button = ctk.CTkButton(root, text='Preview', cursor='hand2', command=lambda: toggle_preview())
92
+ preview_button.place(relx=0.65, rely=0.75, relwidth=0.2, relheight=0.05)
93
+
94
+ status_label = ctk.CTkLabel(root, text=None, justify='center')
95
+ status_label.place(relx=0.1, rely=0.9, relwidth=0.8)
96
+
97
+ donate_label = ctk.CTkLabel(root, text='Become a GitHub Sponsor', justify='center', cursor='hand2')
98
+ donate_label.place(relx=0.1, rely=0.95, relwidth=0.8)
99
+ donate_label.configure(text_color=ctk.ThemeManager.theme.get('RoopDonate').get('text_color'))
100
+ donate_label.bind('<Button>', lambda event: webbrowser.open('https://github.com/sponsors/s0md3v'))
101
+
102
+ return root
103
+
104
+
105
+ def create_preview(parent: ctk.CTkToplevel) -> ctk.CTkToplevel:
106
+ global preview_label, preview_slider
107
+
108
+ preview = ctk.CTkToplevel(parent)
109
+ preview.withdraw()
110
+ preview.title('Preview')
111
+ preview.configure()
112
+ preview.protocol('WM_DELETE_WINDOW', lambda: toggle_preview())
113
+ preview.resizable(width=False, height=False)
114
+
115
+ preview_label = ctk.CTkLabel(preview, text=None)
116
+ preview_label.pack(fill='both', expand=True)
117
+
118
+ preview_slider = ctk.CTkSlider(preview, from_=0, to=0, command=lambda frame_value: update_preview(frame_value))
119
+
120
+ return preview
121
+
122
+
123
+ def update_status(text: str) -> None:
124
+ status_label.configure(text=text)
125
+ ROOT.update()
126
+
127
+
128
+ def select_source_path() -> None:
129
+ global RECENT_DIRECTORY_SOURCE
130
+
131
+ PREVIEW.withdraw()
132
+ source_path = ctk.filedialog.askopenfilename(title='select an source image', initialdir=RECENT_DIRECTORY_SOURCE)
133
+ if is_image(source_path):
134
+ roop.globals.source_path = source_path
135
+ RECENT_DIRECTORY_SOURCE = os.path.dirname(roop.globals.source_path)
136
+ image = render_image_preview(roop.globals.source_path, (200, 200))
137
+ source_label.configure(image=image)
138
+ else:
139
+ roop.globals.source_path = None
140
+ source_label.configure(image=None)
141
+
142
+
143
+ def select_target_path() -> None:
144
+ global RECENT_DIRECTORY_TARGET
145
+
146
+ PREVIEW.withdraw()
147
+ target_path = ctk.filedialog.askopenfilename(title='select an target image or video', initialdir=RECENT_DIRECTORY_TARGET)
148
+ if is_image(target_path):
149
+ roop.globals.target_path = target_path
150
+ RECENT_DIRECTORY_TARGET = os.path.dirname(roop.globals.target_path)
151
+ image = render_image_preview(roop.globals.target_path, (200, 200))
152
+ target_label.configure(image=image)
153
+ elif is_video(target_path):
154
+ roop.globals.target_path = target_path
155
+ RECENT_DIRECTORY_TARGET = os.path.dirname(roop.globals.target_path)
156
+ video_frame = render_video_preview(target_path, (200, 200))
157
+ target_label.configure(image=video_frame)
158
+ else:
159
+ roop.globals.target_path = None
160
+ target_label.configure(image=None)
161
+
162
+
163
+ def select_output_path(start: Callable[[], None]) -> None:
164
+ global RECENT_DIRECTORY_OUTPUT
165
+
166
+ if is_image(roop.globals.target_path):
167
+ output_path = ctk.filedialog.asksaveasfilename(title='save image output file', defaultextension='.png', initialfile='output.png', initialdir=RECENT_DIRECTORY_OUTPUT)
168
+ elif is_video(roop.globals.target_path):
169
+ output_path = ctk.filedialog.asksaveasfilename(title='save video output file', defaultextension='.mp4', initialfile='output.mp4', initialdir=RECENT_DIRECTORY_OUTPUT)
170
+ else:
171
+ output_path = None
172
+ if output_path:
173
+ roop.globals.output_path = output_path
174
+ RECENT_DIRECTORY_OUTPUT = os.path.dirname(roop.globals.output_path)
175
+ start()
176
+
177
+
178
+ def render_image_preview(image_path: str, size: Tuple[int, int]) -> ctk.CTkImage:
179
+ image = Image.open(image_path)
180
+ if size:
181
+ image = ImageOps.fit(image, size, Image.LANCZOS)
182
+ return ctk.CTkImage(image, size=image.size)
183
+
184
+
185
+ def render_video_preview(video_path: str, size: Tuple[int, int], frame_number: int = 0) -> ctk.CTkImage:
186
+ capture = cv2.VideoCapture(video_path)
187
+ if frame_number:
188
+ capture.set(cv2.CAP_PROP_POS_FRAMES, frame_number)
189
+ has_frame, frame = capture.read()
190
+ if has_frame:
191
+ image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
192
+ if size:
193
+ image = ImageOps.fit(image, size, Image.LANCZOS)
194
+ return ctk.CTkImage(image, size=image.size)
195
+ capture.release()
196
+ cv2.destroyAllWindows()
197
+
198
+
199
+ def toggle_preview() -> None:
200
+ if PREVIEW.state() == 'normal':
201
+ PREVIEW.withdraw()
202
+ elif roop.globals.source_path and roop.globals.target_path:
203
+ init_preview()
204
+ update_preview()
205
+ PREVIEW.deiconify()
206
+
207
+
208
+ def init_preview() -> None:
209
+ if is_image(roop.globals.target_path):
210
+ preview_slider.pack_forget()
211
+ if is_video(roop.globals.target_path):
212
+ video_frame_total = get_video_frame_total(roop.globals.target_path)
213
+ preview_slider.configure(to=video_frame_total)
214
+ preview_slider.pack(fill='x')
215
+ preview_slider.set(0)
216
+
217
+
218
+ def update_preview(frame_number: int = 0) -> None:
219
+ if roop.globals.source_path and roop.globals.target_path:
220
+ temp_frame = get_video_frame(roop.globals.target_path, frame_number)
221
+ if predict_frame(temp_frame):
222
+ quit()
223
+ for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
224
+ temp_frame = frame_processor.process_frame(
225
+ get_one_face(cv2.imread(roop.globals.source_path)),
226
+ temp_frame
227
+ )
228
+ image = Image.fromarray(cv2.cvtColor(temp_frame, cv2.COLOR_BGR2RGB))
229
+ image = ImageOps.contain(image, (PREVIEW_MAX_WIDTH, PREVIEW_MAX_HEIGHT), Image.LANCZOS)
230
+ image = ctk.CTkImage(image, size=image.size)
231
+ preview_label.configure(image=image)
roop/utilities.py ADDED
@@ -0,0 +1,141 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import glob
2
+ import mimetypes
3
+ import os
4
+ import platform
5
+ import shutil
6
+ import ssl
7
+ import subprocess
8
+ import urllib
9
+ from pathlib import Path
10
+ from typing import List, Any
11
+ from tqdm import tqdm
12
+
13
+ import roop.globals
14
+
15
+ TEMP_FILE = 'temp.mp4'
16
+ TEMP_DIRECTORY = 'temp'
17
+
18
+ # monkey patch ssl for mac
19
+ if platform.system().lower() == 'darwin':
20
+ ssl._create_default_https_context = ssl._create_unverified_context
21
+
22
+
23
+ def run_ffmpeg(args: List[str]) -> bool:
24
+ commands = ['ffmpeg', '-hide_banner', '-hwaccel', 'auto', '-loglevel', roop.globals.log_level]
25
+ commands.extend(args)
26
+ try:
27
+ subprocess.check_output(commands, stderr=subprocess.STDOUT)
28
+ return True
29
+ except Exception:
30
+ pass
31
+ return False
32
+
33
+
34
+ def detect_fps(target_path: str) -> float:
35
+ command = ['ffprobe', '-v', 'error', '-select_streams', 'v:0', '-show_entries', 'stream=r_frame_rate', '-of', 'default=noprint_wrappers=1:nokey=1', target_path]
36
+ output = subprocess.check_output(command).decode().strip().split('/')
37
+ try:
38
+ numerator, denominator = map(int, output)
39
+ return numerator / denominator
40
+ except Exception:
41
+ pass
42
+ return 30.0
43
+
44
+
45
+ def extract_frames(target_path: str) -> None:
46
+ temp_directory_path = get_temp_directory_path(target_path)
47
+ run_ffmpeg(['-i', target_path, '-pix_fmt', 'rgb24', os.path.join(temp_directory_path, '%04d.png')])
48
+
49
+
50
+ def create_video(target_path: str, fps: float = 30.0) -> None:
51
+ temp_output_path = get_temp_output_path(target_path)
52
+ temp_directory_path = get_temp_directory_path(target_path)
53
+ run_ffmpeg(['-r', str(fps), '-i', os.path.join(temp_directory_path, '%04d.png'), '-c:v', roop.globals.video_encoder, '-crf', str(roop.globals.video_quality), '-pix_fmt', 'yuv420p', '-vf', 'colorspace=bt709:iall=bt601-6-625:fast=1', '-y', temp_output_path])
54
+
55
+
56
+ def restore_audio(target_path: str, output_path: str) -> None:
57
+ temp_output_path = get_temp_output_path(target_path)
58
+ done = run_ffmpeg(['-i', temp_output_path, '-i', target_path, '-c:v', 'copy', '-map', '0:v:0', '-map', '1:a:0', '-y', output_path])
59
+ if not done:
60
+ move_temp(target_path, output_path)
61
+
62
+
63
+ def get_temp_frame_paths(target_path: str) -> List[str]:
64
+ temp_directory_path = get_temp_directory_path(target_path)
65
+ return glob.glob((os.path.join(glob.escape(temp_directory_path), '*.png')))
66
+
67
+
68
+ def get_temp_directory_path(target_path: str) -> str:
69
+ target_name, _ = os.path.splitext(os.path.basename(target_path))
70
+ target_directory_path = os.path.dirname(target_path)
71
+ return os.path.join(target_directory_path, TEMP_DIRECTORY, target_name)
72
+
73
+
74
+ def get_temp_output_path(target_path: str) -> str:
75
+ temp_directory_path = get_temp_directory_path(target_path)
76
+ return os.path.join(temp_directory_path, TEMP_FILE)
77
+
78
+
79
+ def normalize_output_path(source_path: str, target_path: str, output_path: str) -> Any:
80
+ if source_path and target_path:
81
+ source_name, _ = os.path.splitext(os.path.basename(source_path))
82
+ target_name, target_extension = os.path.splitext(os.path.basename(target_path))
83
+ if os.path.isdir(output_path):
84
+ return os.path.join(output_path, source_name + '-' + target_name + target_extension)
85
+ return output_path
86
+
87
+
88
+ def create_temp(target_path: str) -> None:
89
+ temp_directory_path = get_temp_directory_path(target_path)
90
+ Path(temp_directory_path).mkdir(parents=True, exist_ok=True)
91
+
92
+
93
+ def move_temp(target_path: str, output_path: str) -> None:
94
+ temp_output_path = get_temp_output_path(target_path)
95
+ if os.path.isfile(temp_output_path):
96
+ if os.path.isfile(output_path):
97
+ os.remove(output_path)
98
+ shutil.move(temp_output_path, output_path)
99
+
100
+
101
+ def clean_temp(target_path: str) -> None:
102
+ temp_directory_path = get_temp_directory_path(target_path)
103
+ parent_directory_path = os.path.dirname(temp_directory_path)
104
+ if not roop.globals.keep_frames and os.path.isdir(temp_directory_path):
105
+ shutil.rmtree(temp_directory_path)
106
+ if os.path.exists(parent_directory_path) and not os.listdir(parent_directory_path):
107
+ os.rmdir(parent_directory_path)
108
+
109
+
110
+ def has_image_extension(image_path: str) -> bool:
111
+ return image_path.lower().endswith(('png', 'jpg', 'jpeg'))
112
+
113
+
114
+ def is_image(image_path: str) -> bool:
115
+ if image_path and os.path.isfile(image_path):
116
+ mimetype, _ = mimetypes.guess_type(image_path)
117
+ return bool(mimetype and mimetype.startswith('image/'))
118
+ return False
119
+
120
+
121
+ def is_video(video_path: str) -> bool:
122
+ if video_path and os.path.isfile(video_path):
123
+ mimetype, _ = mimetypes.guess_type(video_path)
124
+ return bool(mimetype and mimetype.startswith('video/'))
125
+ return False
126
+
127
+
128
+ def conditional_download(download_directory_path: str, urls: List[str]) -> None:
129
+ if not os.path.exists(download_directory_path):
130
+ os.makedirs(download_directory_path)
131
+ for url in urls:
132
+ download_file_path = os.path.join(download_directory_path, os.path.basename(url))
133
+ if not os.path.exists(download_file_path):
134
+ request = urllib.request.urlopen(url) # type: ignore[attr-defined]
135
+ total = int(request.headers.get('Content-Length', 0))
136
+ with tqdm(total=total, desc='Downloading', unit='B', unit_scale=True, unit_divisor=1024) as progress:
137
+ urllib.request.urlretrieve(url, download_file_path, reporthook=lambda count, block_size, total_size: progress.update(block_size)) # type: ignore[attr-defined]
138
+
139
+
140
+ def resolve_relative_path(path: str) -> str:
141
+ return os.path.abspath(os.path.join(os.path.dirname(__file__), path))