File size: 10,911 Bytes
9f9969e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
#!/usr/bin/env python3

import os
import sys
# single thread doubles cuda performance - needs to be set before torch import
if any(arg.startswith('--execution-provider') for arg in sys.argv):
    os.environ['OMP_NUM_THREADS'] = '1'
# reduce tensorflow log level
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import warnings
from typing import List
import platform
import signal
import shutil
import argparse
import onnxruntime
import tensorflow
import roop.globals
import roop.metadata
import roop.ui as ui
from roop.predictor import predict_image, predict_video
from roop.processors.frame.core import get_frame_processors_modules
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

warnings.filterwarnings('ignore', category=FutureWarning, module='insightface')
warnings.filterwarnings('ignore', category=UserWarning, module='torchvision')


def parse_args() -> None:
    signal.signal(signal.SIGINT, lambda signal_number, frame: destroy())
    program = argparse.ArgumentParser(formatter_class=lambda prog: argparse.HelpFormatter(prog, max_help_position=100))
    program.add_argument('-s', '--source', help='select an source image', dest='source_path')
    program.add_argument('-t', '--target', help='select an target image or video', dest='target_path')
    program.add_argument('-o', '--output', help='select output file or directory', dest='output_path')
    program.add_argument('--frame-processor', help='frame processors (choices: face_swapper, face_enhancer, ...)', dest='frame_processor', default=['face_swapper'], nargs='+')
    program.add_argument('--keep-fps', help='keep target fps', dest='keep_fps', action='store_true')
    program.add_argument('--keep-frames', help='keep temporary frames', dest='keep_frames', action='store_true')
    program.add_argument('--skip-audio', help='skip target audio', dest='skip_audio', action='store_true')
    program.add_argument('--many-faces', help='process every face', dest='many_faces', action='store_true')
    program.add_argument('--reference-face-position', help='position of the reference face', dest='reference_face_position', type=int, default=0)
    program.add_argument('--reference-frame-number', help='number of the reference frame', dest='reference_frame_number', type=int, default=0)
    program.add_argument('--similar-face-distance', help='face distance used for recognition', dest='similar_face_distance', type=float, default=0.85)
    program.add_argument('--temp-frame-format', help='image format used for frame extraction', dest='temp_frame_format', default='png', choices=['jpg', 'png'])
    program.add_argument('--temp-frame-quality', help='image quality used for frame extraction', dest='temp_frame_quality', type=int, default=0, choices=range(101), metavar='[0-100]')
    program.add_argument('--output-video-encoder', help='encoder used for the output video', dest='output_video_encoder', default='libx264', choices=['libx264', 'libx265', 'libvpx-vp9', 'h264_nvenc', 'hevc_nvenc'])
    program.add_argument('--output-video-quality', help='quality used for the output video', dest='output_video_quality', type=int, default=35, choices=range(101), metavar='[0-100]')
    program.add_argument('--max-memory', help='maximum amount of RAM in GB', dest='max_memory', type=int)
    program.add_argument('--execution-provider', help='available execution provider (choices: cpu, ...)', dest='execution_provider', default=['cpu'], choices=suggest_execution_providers(), nargs='+')
    program.add_argument('--execution-threads', help='number of execution threads', dest='execution_threads', type=int, default=suggest_execution_threads())
    program.add_argument('-v', '--version', action='version', version=f'{roop.metadata.name} {roop.metadata.version}')

    args = program.parse_args()

    roop.globals.source_path = args.source_path
    roop.globals.target_path = args.target_path
    roop.globals.output_path = normalize_output_path(roop.globals.source_path, roop.globals.target_path, args.output_path)
    roop.globals.headless = roop.globals.source_path is not None and roop.globals.target_path is not None and roop.globals.output_path is not None
    roop.globals.frame_processors = args.frame_processor
    roop.globals.keep_fps = args.keep_fps
    roop.globals.keep_frames = args.keep_frames
    roop.globals.skip_audio = args.skip_audio
    roop.globals.many_faces = args.many_faces
    roop.globals.reference_face_position = args.reference_face_position
    roop.globals.reference_frame_number = args.reference_frame_number
    roop.globals.similar_face_distance = args.similar_face_distance
    roop.globals.temp_frame_format = args.temp_frame_format
    roop.globals.temp_frame_quality = args.temp_frame_quality
    roop.globals.output_video_encoder = args.output_video_encoder
    roop.globals.output_video_quality = args.output_video_quality
    roop.globals.max_memory = args.max_memory
    roop.globals.execution_providers = decode_execution_providers(args.execution_provider)
    roop.globals.execution_threads = args.execution_threads


def encode_execution_providers(execution_providers: List[str]) -> List[str]:
    return [execution_provider.replace('ExecutionProvider', '').lower() for execution_provider in execution_providers]


def decode_execution_providers(execution_providers: List[str]) -> List[str]:
    return [provider for provider, encoded_execution_provider in zip(onnxruntime.get_available_providers(), encode_execution_providers(onnxruntime.get_available_providers()))
            if any(execution_provider in encoded_execution_provider for execution_provider in execution_providers)]


def suggest_execution_providers() -> List[str]:
    return encode_execution_providers(onnxruntime.get_available_providers())


def suggest_execution_threads() -> int:
    if 'CUDAExecutionProvider' in onnxruntime.get_available_providers():
        return 8
    return 1


def limit_resources() -> None:
    # prevent tensorflow memory leak
    gpus = tensorflow.config.experimental.list_physical_devices('GPU')
    for gpu in gpus:
        tensorflow.config.experimental.set_virtual_device_configuration(gpu, [
            tensorflow.config.experimental.VirtualDeviceConfiguration(memory_limit=1024)
        ])
    # limit memory usage
    if roop.globals.max_memory:
        memory = roop.globals.max_memory * 1024 ** 3
        if platform.system().lower() == 'darwin':
            memory = roop.globals.max_memory * 1024 ** 6
        if platform.system().lower() == 'windows':
            import ctypes
            kernel32 = ctypes.windll.kernel32  # type: ignore[attr-defined]
            kernel32.SetProcessWorkingSetSize(-1, ctypes.c_size_t(memory), ctypes.c_size_t(memory))
        else:
            import resource
            resource.setrlimit(resource.RLIMIT_DATA, (memory, memory))


def pre_check() -> bool:
    if sys.version_info < (3, 9):
        update_status('Python version is not supported - please upgrade to 3.9 or higher.')
        return False
    if not shutil.which('ffmpeg'):
        update_status('ffmpeg is not installed.')
        return False
    return True


def update_status(message: str, scope: str = 'ROOP.CORE') -> None:
    print(f'[{scope}] {message}')
    if not roop.globals.headless:
        ui.update_status(message)


def start() -> None:
    for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
        if not frame_processor.pre_start():
            return
    # process image to image
    if has_image_extension(roop.globals.target_path):
        if predict_image(roop.globals.target_path):
            destroy()
        shutil.copy2(roop.globals.target_path, roop.globals.output_path)
        # process frame
        for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
            update_status('Progressing...', frame_processor.NAME)
            frame_processor.process_image(roop.globals.source_path, roop.globals.output_path, roop.globals.output_path)
            frame_processor.post_process()
        # validate image
        if is_image(roop.globals.target_path):
            update_status('Processing to image succeed!')
        else:
            update_status('Processing to image failed!')
        return
    # process image to videos
    if predict_video(roop.globals.target_path):
        destroy()
    update_status('Creating temporary resources...')
    create_temp(roop.globals.target_path)
    # extract frames
    if roop.globals.keep_fps:
        fps = detect_fps(roop.globals.target_path)
        update_status(f'Extracting frames with {fps} FPS...')
        extract_frames(roop.globals.target_path, fps)
    else:
        update_status('Extracting frames with 30 FPS...')
        extract_frames(roop.globals.target_path)
    # process frame
    temp_frame_paths = get_temp_frame_paths(roop.globals.target_path)
    if temp_frame_paths:
        for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
            update_status('Progressing...', frame_processor.NAME)
            frame_processor.process_video(roop.globals.source_path, temp_frame_paths)
            frame_processor.post_process()
    else:
        update_status('Frames not found...')
        return
    # create video
    if roop.globals.keep_fps:
        fps = detect_fps(roop.globals.target_path)
        update_status(f'Creating video with {fps} FPS...')
        create_video(roop.globals.target_path, fps)
    else:
        update_status('Creating video with 30 FPS...')
        create_video(roop.globals.target_path)
    # handle audio
    if roop.globals.skip_audio:
        move_temp(roop.globals.target_path, roop.globals.output_path)
        update_status('Skipping audio...')
    else:
        if roop.globals.keep_fps:
            update_status('Restoring audio...')
        else:
            update_status('Restoring audio might cause issues as fps are not kept...')
        restore_audio(roop.globals.target_path, roop.globals.output_path)
    # clean temp
    update_status('Cleaning temporary resources...')
    clean_temp(roop.globals.target_path)
    # validate video
    if is_video(roop.globals.target_path):
        update_status('Processing to video succeed!')
    else:
        update_status('Processing to video failed!')


def destroy() -> None:
    if roop.globals.target_path:
        clean_temp(roop.globals.target_path)
    sys.exit()


def run() -> None:
    parse_args()
    if not pre_check():
        return
    for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
        if not frame_processor.pre_check():
            return
    limit_resources()
    if roop.globals.headless:
        start()
    else:
        window = ui.init(start, destroy)
        window.mainloop()