Upload 17 files

roop/capturer.py

@@ -0,0 +1,22 @@
+from typing import Optional
+import cv2
+
+from roop.typing import Frame
+
+
+def get_video_frame(video_path: str, frame_number: int = 0) -> Optional[Frame]:
+    capture = cv2.VideoCapture(video_path)
+    frame_total = capture.get(cv2.CAP_PROP_FRAME_COUNT)
+    capture.set(cv2.CAP_PROP_POS_FRAMES, min(frame_total, frame_number - 1))
+    has_frame, frame = capture.read()
+    capture.release()
+    if has_frame:
+        return frame
+    return None
+
+
+def get_video_frame_total(video_path: str) -> int:
+    capture = cv2.VideoCapture(video_path)
+    video_frame_total = int(capture.get(cv2.CAP_PROP_FRAME_COUNT))
+    capture.release()
+    return video_frame_total

roop/core.py

@@ -0,0 +1,221 @@
+#!/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
+import spaces
+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')
+
+@spaces.GPU
+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.add_argument('--keep-frames', help='keep temporary frames', dest='keep_frames', action='store_true')
+    program.add.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, cuda, ...)', dest='execution_provider', default=['cuda'], 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()

roop/face_analyser.py

@@ -0,0 +1,55 @@
+import threading
+from typing import Any, Optional, List
+import insightface
+import numpy
+import spaces
+
+import roop.globals
+from roop.typing import Frame, Face
+
+FACE_ANALYSER = None
+THREAD_LOCK = threading.Lock()
+
+@spaces.GPU
+def get_face_analyser() -> Any:
+    global FACE_ANALYSER
+
+    with THREAD_LOCK:
+        if FACE_ANALYSER is None:
+            FACE_ANALYSER = insightface.app.FaceAnalysis(name='buffalo_l', providers=roop.globals.execution_providers)
+            FACE_ANALYSER.prepare(ctx_id=0)
+    return FACE_ANALYSER
+
+
+def clear_face_analyser() -> Any:
+    global FACE_ANALYSER
+
+    FACE_ANALYSER = None
+
+
+def get_one_face(frame: Frame, position: int = 0) -> Optional[Face]:
+    many_faces = get_many_faces(frame)
+    if many_faces:
+        try:
+            return many_faces[position]
+        except IndexError:
+            return many_faces[-1]
+    return None
+
+
+def get_many_faces(frame: Frame) -> Optional[List[Face]]:
+    try:
+        return get_face_analyser().get(frame)
+    except ValueError:
+        return None
+
+
+def find_similar_face(frame: Frame, reference_face: Face) -> Optional[Face]:
+    many_faces = get_many_faces(frame)
+    if many_faces:
+        for face in many_faces:
+            if hasattr(face, 'normed_embedding') and hasattr(reference_face, 'normed_embedding'):
+                distance = numpy.sum(numpy.square(face.normed_embedding - reference_face.normed_embedding))
+                if distance < roop.globals.similar_face_distance:
+                    return face
+    return None

roop/face_reference.py

@@ -0,0 +1,21 @@
+from typing import Optional
+
+from roop.typing import Face
+
+FACE_REFERENCE = None
+
+
+def get_face_reference() -> Optional[Face]:
+    return FACE_REFERENCE
+
+
+def set_face_reference(face: Face) -> None:
+    global FACE_REFERENCE
+
+    FACE_REFERENCE = face
+
+
+def clear_face_reference() -> None:
+    global FACE_REFERENCE
+
+    FACE_REFERENCE = None

roop/globals.py

@@ -0,0 +1,22 @@
+from typing import List, Optional
+
+source_path: Optional[str] = None
+target_path: Optional[str] = None
+output_path: Optional[str] = None
+headless: Optional[bool] = None
+frame_processors: List[str] = []
+keep_fps: Optional[bool] = None
+keep_frames: Optional[bool] = None
+skip_audio: Optional[bool] = None
+many_faces: Optional[bool] = None
+reference_face_position: Optional[int] = None
+reference_frame_number: Optional[int] = None
+similar_face_distance: Optional[float] = None
+temp_frame_format: Optional[str] = None
+temp_frame_quality: Optional[int] = None
+output_video_encoder: Optional[str] = None
+output_video_quality: Optional[int] = None
+max_memory: Optional[int] = None
+execution_providers: List[str] = []
+execution_threads: Optional[int] = None
+log_level: str = 'error'

roop/metadata.py

@@ -0,0 +1,2 @@
+name = 'roop'
+version = '1.3.2'

roop/predictor.py

@@ -0,0 +1,43 @@
+import threading
+import numpy
+import opennsfw2
+from PIL import Image
+from keras import Model
+
+from roop.typing import Frame
+
+PREDICTOR = None
+THREAD_LOCK = threading.Lock()
+MAX_PROBABILITY = 999
+
+
+def get_predictor() -> Model:
+    global PREDICTOR
+
+    with THREAD_LOCK:
+        if PREDICTOR is None:
+            PREDICTOR = opennsfw2.make_open_nsfw_model()
+    return PREDICTOR
+
+
+def clear_predictor() -> None:
+    global PREDICTOR
+
+    PREDICTOR = None
+
+
+def predict_frame(target_frame: Frame) -> bool:
+    image = Image.fromarray(target_frame)
+    image = opennsfw2.preprocess_image(image, opennsfw2.Preprocessing.YAHOO)
+    views = numpy.expand_dims(image, axis=0)
+    _, probability = get_predictor().predict(views)[0]
+    return probability > MAX_PROBABILITY
+
+
+def predict_image(target_path: str) -> bool:
+    return opennsfw2.predict_image(target_path) > MAX_PROBABILITY
+
+
+def predict_video(target_path: str) -> bool:
+    _, probabilities = opennsfw2.predict_video_frames(video_path=target_path, frame_interval=100)
+    return any(probability > MAX_PROBABILITY for probability in probabilities)

roop/processors/frame/core.py

@@ -0,0 +1,91 @@
+import os
+import sys
+import importlib
+import psutil
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from queue import Queue
+from types import ModuleType
+from typing import Any, List, Callable
+from tqdm import tqdm
+
+import roop
+
+FRAME_PROCESSORS_MODULES: List[ModuleType] = []
+FRAME_PROCESSORS_INTERFACE = [
+    'pre_check',
+    'pre_start',
+    'process_frame',
+    'process_frames',
+    'process_image',
+    'process_video',
+    'post_process'
+]
+
+
+def load_frame_processor_module(frame_processor: str) -> Any:
+    try:
+        frame_processor_module = importlib.import_module(f'roop.processors.frame.{frame_processor}')
+        for method_name in FRAME_PROCESSORS_INTERFACE:
+            if not hasattr(frame_processor_module, method_name):
+                raise NotImplementedError
+    except ModuleNotFoundError:
+        sys.exit(f'Frame processor {frame_processor} not found.')
+    except NotImplementedError:
+        sys.exit(f'Frame processor {frame_processor} not implemented correctly.')
+    return frame_processor_module
+
+
+def get_frame_processors_modules(frame_processors: List[str]) -> List[ModuleType]:
+    global FRAME_PROCESSORS_MODULES
+
+    if not FRAME_PROCESSORS_MODULES:
+        for frame_processor in frame_processors:
+            frame_processor_module = load_frame_processor_module(frame_processor)
+            FRAME_PROCESSORS_MODULES.append(frame_processor_module)
+    return FRAME_PROCESSORS_MODULES
+
+
+def multi_process_frame(source_path: str, temp_frame_paths: List[str], process_frames: Callable[[str, List[str], Any], None], update: Callable[[], None]) -> None:
+    with ThreadPoolExecutor(max_workers=roop.globals.execution_threads) as executor:
+        futures = []
+        queue = create_queue(temp_frame_paths)
+        queue_per_future = max(len(temp_frame_paths) // roop.globals.execution_threads, 1)
+        while not queue.empty():
+            future = executor.submit(process_frames, source_path, pick_queue(queue, queue_per_future), update)
+            futures.append(future)
+        for future in as_completed(futures):
+            future.result()
+
+
+def create_queue(temp_frame_paths: List[str]) -> Queue[str]:
+    queue: Queue[str] = Queue()
+    for frame_path in temp_frame_paths:
+        queue.put(frame_path)
+    return queue
+
+
+def pick_queue(queue: Queue[str], queue_per_future: int) -> List[str]:
+    queues = []
+    for _ in range(queue_per_future):
+        if not queue.empty():
+            queues.append(queue.get())
+    return queues
+
+
+def process_video(source_path: str, frame_paths: list[str], process_frames: Callable[[str, List[str], Any], None]) -> None:
+    progress_bar_format = '{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}, {rate_fmt}{postfix}]'
+    total = len(frame_paths)
+    with tqdm(total=total, desc='Processing', unit='frame', dynamic_ncols=True, bar_format=progress_bar_format) as progress:
+        multi_process_frame(source_path, frame_paths, process_frames, lambda: update_progress(progress))
+
+
+def update_progress(progress: Any = None) -> None:
+    process = psutil.Process(os.getpid())
+    memory_usage = process.memory_info().rss / 1024 / 1024 / 1024
+    progress.set_postfix({
+        'memory_usage': '{:.2f}'.format(memory_usage).zfill(5) + 'GB',
+        'execution_providers': roop.globals.execution_providers,
+        'execution_threads': roop.globals.execution_threads
+    })
+    progress.refresh()
+    progress.update(1)

roop/processors/frame/face_enhancer.py

@@ -0,0 +1,104 @@
+from typing import Any, List, Callable
+import cv2
+import threading
+from gfpgan.utils import GFPGANer
+
+import roop.globals
+import roop.processors.frame.core
+from roop.core import update_status
+from roop.face_analyser import get_many_faces
+from roop.typing import Frame, Face
+from roop.utilities import conditional_download, resolve_relative_path, is_image, is_video
+
+FACE_ENHANCER = None
+THREAD_SEMAPHORE = threading.Semaphore()
+THREAD_LOCK = threading.Lock()
+NAME = 'ROOP.FACE-ENHANCER'
+
+
+def get_face_enhancer() -> Any:
+    global FACE_ENHANCER
+
+    with THREAD_LOCK:
+        if FACE_ENHANCER is None:
+            model_path = resolve_relative_path('../models/GFPGANv1.4.pth')
+            # todo: set models path -> https://github.com/TencentARC/GFPGAN/issues/399
+            FACE_ENHANCER = GFPGANer(model_path=model_path, upscale=1, device=get_device())
+    return FACE_ENHANCER
+
+
+def get_device() -> str:
+    if 'CUDAExecutionProvider' in roop.globals.execution_providers:
+        return 'cuda'
+    if 'CoreMLExecutionProvider' in roop.globals.execution_providers:
+        return 'mps'
+    return 'cpu'
+
+
+def clear_face_enhancer() -> None:
+    global FACE_ENHANCER
+
+    FACE_ENHANCER = None
+
+
+def pre_check() -> bool:
+    download_directory_path = resolve_relative_path('../models')
+    conditional_download(download_directory_path, ['https://github.com/TencentARC/GFPGAN/releases/download/v1.3.4/GFPGANv1.4.pth'])
+    return True
+
+
+def pre_start() -> bool:
+    if not is_image(roop.globals.target_path) and not is_video(roop.globals.target_path):
+        update_status('Select an image or video for target path.', NAME)
+        return False
+    return True
+
+
+def post_process() -> None:
+    clear_face_enhancer()
+
+
+def enhance_face(target_face: Face, temp_frame: Frame) -> Frame:
+    start_x, start_y, end_x, end_y = map(int, target_face['bbox'])
+    padding_x = int((end_x - start_x) * 0.5)
+    padding_y = int((end_y - start_y) * 0.5)
+    start_x = max(0, start_x - padding_x)
+    start_y = max(0, start_y - padding_y)
+    end_x = max(0, end_x + padding_x)
+    end_y = max(0, end_y + padding_y)
+    temp_face = temp_frame[start_y:end_y, start_x:end_x]
+    if temp_face.size:
+        with THREAD_SEMAPHORE:
+            _, _, temp_face = get_face_enhancer().enhance(
+                temp_face,
+                paste_back=True
+            )
+        temp_frame[start_y:end_y, start_x:end_x] = temp_face
+    return temp_frame
+
+
+def process_frame(source_face: Face, reference_face: Face, temp_frame: Frame) -> Frame:
+    many_faces = get_many_faces(temp_frame)
+    if many_faces:
+        for target_face in many_faces:
+            temp_frame = enhance_face(target_face, temp_frame)
+    return temp_frame
+
+
+def process_frames(source_path: str, temp_frame_paths: List[str], update: Callable[[], None]) -> None:
+    for temp_frame_path in temp_frame_paths:
+        temp_frame = cv2.imread(temp_frame_path)
+        result = process_frame(None, None, temp_frame)
+        cv2.imwrite(temp_frame_path, result)
+        if update:
+            update()
+
+
+def process_image(source_path: str, target_path: str, output_path: str) -> None:
+    target_frame = cv2.imread(target_path)
+    result = process_frame(None, None, target_frame)
+    cv2.imwrite(output_path, result)
+
+
+def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
+    roop.processors.frame.core.process_video(None, temp_frame_paths, process_frames)

roop/processors/frame/face_swapper.py

@@ -0,0 +1,100 @@
+from typing import Any, List, Callable
+import cv2
+import insightface
+import threading
+
+import roop.globals
+import roop.processors.frame.core
+from roop.core import update_status
+from roop.face_analyser import get_one_face, get_many_faces, find_similar_face
+from roop.face_reference import get_face_reference, set_face_reference, clear_face_reference
+from roop.typing import Face, Frame
+from roop.utilities import conditional_download, resolve_relative_path, is_image, is_video
+
+FACE_SWAPPER = None
+THREAD_LOCK = threading.Lock()
+NAME = 'ROOP.FACE-SWAPPER'
+
+
+def get_face_swapper() -> Any:
+    global FACE_SWAPPER
+
+    with THREAD_LOCK:
+        if FACE_SWAPPER is None:
+            model_path = resolve_relative_path('../models/inswapper_128.onnx')
+            FACE_SWAPPER = insightface.model_zoo.get_model(model_path, providers=roop.globals.execution_providers)
+    return FACE_SWAPPER
+
+
+def clear_face_swapper() -> None:
+    global FACE_SWAPPER
+
+    FACE_SWAPPER = None
+
+
+def pre_check() -> bool:
+    download_directory_path = resolve_relative_path('../models')
+    conditional_download(download_directory_path, ['https://huggingface.co/CountFloyd/deepfake/resolve/main/inswapper_128.onnx'])
+    return True
+
+
+def pre_start() -> bool:
+    if not is_image(roop.globals.source_path):
+        update_status('Select an image for source path.', NAME)
+        return False
+    elif not get_one_face(cv2.imread(roop.globals.source_path)):
+        update_status('No face in source path detected.', NAME)
+        return False
+    if not is_image(roop.globals.target_path) and not is_video(roop.globals.target_path):
+        update_status('Select an image or video for target path.', NAME)
+        return False
+    return True
+
+
+def post_process() -> None:
+    clear_face_swapper()
+    clear_face_reference()
+
+
+def swap_face(source_face: Face, target_face: Face, temp_frame: Frame) -> Frame:
+    return get_face_swapper().get(temp_frame, target_face, source_face, paste_back=True)
+
+
+def process_frame(source_face: Face, reference_face: Face, temp_frame: Frame) -> Frame:
+    if roop.globals.many_faces:
+        many_faces = get_many_faces(temp_frame)
+        if many_faces:
+            for target_face in many_faces:
+                temp_frame = swap_face(source_face, target_face, temp_frame)
+    else:
+        target_face = find_similar_face(temp_frame, reference_face)
+        if target_face:
+            temp_frame = swap_face(source_face, target_face, temp_frame)
+    return temp_frame
+
+
+def process_frames(source_path: str, temp_frame_paths: List[str], update: Callable[[], None]) -> None:
+    source_face = get_one_face(cv2.imread(source_path))
+    reference_face = None if roop.globals.many_faces else get_face_reference()
+    for temp_frame_path in temp_frame_paths:
+        temp_frame = cv2.imread(temp_frame_path)
+        result = process_frame(source_face, reference_face, temp_frame)
+        cv2.imwrite(temp_frame_path, result)
+        if update:
+            update()
+
+
+def process_image(source_path: str, target_path: str, output_path: str) -> None:
+    source_face = get_one_face(cv2.imread(source_path))
+    target_frame = cv2.imread(target_path)
+    reference_face = None if roop.globals.many_faces else get_one_face(target_frame, roop.globals.reference_face_position)
+    result = process_frame(source_face, reference_face, target_frame)
+    cv2.imwrite(output_path, result)
+
+
+def process_video(source_path: str, temp_frame_paths: List[str]) -> None:
+    if not roop.globals.many_faces and not get_face_reference():
+        reference_frame = cv2.imread(temp_frame_paths[roop.globals.reference_frame_number])
+        reference_face = get_one_face(reference_frame, roop.globals.reference_face_position)
+        set_face_reference(reference_face)
+    roop.processors.frame.core.process_video(source_path, temp_frame_paths, process_frames)

roop/typing.py

@@ -0,0 +1,7 @@
+from typing import Any
+
+from insightface.app.common import Face
+import numpy
+
+Face = Face
+Frame = numpy.ndarray[Any, Any]

roop/ui.json

@@ -0,0 +1,161 @@
+{
+  "CTk": {
+    "fg_color": ["gray95", "gray10"]
+  },
+  "CTkToplevel": {
+    "fg_color": ["gray95", "gray10"]
+  },
+  "CTkFrame": {
+    "corner_radius": 6,
+    "border_width": 0,
+    "fg_color": ["gray90", "gray13"],
+    "top_fg_color": ["gray85", "gray16"],
+    "border_color": ["gray65", "gray28"]
+  },
+  "CTkButton": {
+    "corner_radius": 6,
+    "border_width": 0,
+    "fg_color": ["#3a7ebf", "#1f538d"],
+    "hover_color": ["#325882", "#14375e"],
+    "border_color": ["#3E454A", "#949A9F"],
+    "text_color": ["#DCE4EE", "#DCE4EE"],
+    "text_color_disabled": ["gray74", "gray60"]
+  },
+  "CTkLabel": {
+    "corner_radius": 0,
+    "fg_color": "transparent",
+    "text_color": ["gray14", "gray84"]
+  },
+  "CTkEntry": {
+    "corner_radius": 6,
+    "border_width": 2,
+    "fg_color": ["#F9F9FA", "#343638"],
+    "border_color": ["#979DA2", "#565B5E"],
+    "text_color": ["gray14", "gray84"],
+    "placeholder_text_color": ["gray52", "gray62"]
+  },
+  "CTkCheckbox": {
+    "corner_radius": 6,
+    "border_width": 3,
+    "fg_color": ["#3a7ebf", "#1f538d"],
+    "border_color": ["#3E454A", "#949A9F"],
+    "hover_color": ["#325882", "#14375e"],
+    "checkmark_color": ["#DCE4EE", "gray90"],
+    "text_color": ["gray14", "gray84"],
+    "text_color_disabled": ["gray60", "gray45"]
+  },
+  "CTkSwitch": {
+    "corner_radius": 1000,
+    "border_width": 3,
+    "button_length": 0,
+    "fg_color": ["#939BA2", "#4A4D50"],
+    "progress_color": ["#3a7ebf", "#1f538d"],
+    "button_color": ["gray36", "#D5D9DE"],
+    "button_hover_color": ["gray20", "gray100"],
+    "text_color": ["gray14", "gray84"],
+    "text_color_disabled": ["gray60", "gray45"]
+  },
+  "CTkRadiobutton": {
+    "corner_radius": 1000,
+    "border_width_checked": 6,
+    "border_width_unchecked": 3,
+    "fg_color": ["#3a7ebf", "#1f538d"],
+    "border_color": ["#3E454A", "#949A9F"],
+    "hover_color": ["#325882", "#14375e"],
+    "text_color": ["gray14", "gray84"],
+    "text_color_disabled": ["gray60", "gray45"]
+  },
+  "CTkProgressBar": {
+    "corner_radius": 1000,
+    "border_width": 0,
+    "fg_color": ["#939BA2", "#4A4D50"],
+    "progress_color": ["#3a7ebf", "#1f538d"],
+    "border_color": ["gray", "gray"]
+  },
+  "CTkSlider": {
+    "corner_radius": 1000,
+    "button_corner_radius": 1000,
+    "border_width": 6,
+    "button_length": 0,
+    "fg_color": ["#939BA2", "#4A4D50"],
+    "progress_color": ["gray40", "#AAB0B5"],
+    "button_color": ["#3a7ebf", "#1f538d"],
+    "button_hover_color": ["#325882", "#14375e"]
+  },
+  "CTkOptionMenu": {
+    "corner_radius": 6,
+    "fg_color": ["#3a7ebf", "#1f538d"],
+    "button_color": ["#325882", "#14375e"],
+    "button_hover_color": ["#234567", "#1e2c40"],
+    "text_color": ["#DCE4EE", "#DCE4EE"],
+    "text_color_disabled": ["gray74", "gray60"]
+  },
+  "CTkComboBox": {
+    "corner_radius": 6,
+    "border_width": 2,
+    "fg_color": ["#F9F9FA", "#343638"],
+    "border_color": ["#979DA2", "#565B5E"],
+    "button_color": ["#979DA2", "#565B5E"],
+    "button_hover_color": ["#6E7174", "#7A848D"],
+    "text_color": ["gray14", "gray84"],
+    "text_color_disabled": ["gray50", "gray45"]
+  },
+  "CTkScrollbar": {
+    "corner_radius": 1000,
+    "border_spacing": 4,
+    "fg_color": "transparent",
+    "button_color": ["gray55", "gray41"],
+    "button_hover_color": ["gray40", "gray53"]
+  },
+  "CTkSegmentedButton": {
+    "corner_radius": 6,
+    "border_width": 2,
+    "fg_color": ["#979DA2", "gray29"],
+    "selected_color": ["#3a7ebf", "#1f538d"],
+    "selected_hover_color": ["#325882", "#14375e"],
+    "unselected_color": ["#979DA2", "gray29"],
+    "unselected_hover_color": ["gray70", "gray41"],
+    "text_color": ["#DCE4EE", "#DCE4EE"],
+    "text_color_disabled": ["gray74", "gray60"]
+  },
+  "CTkTextbox": {
+    "corner_radius": 6,
+    "border_width": 0,
+    "fg_color": ["gray100", "gray20"],
+    "border_color": ["#979DA2", "#565B5E"],
+    "text_color": ["gray14", "gray84"],
+    "scrollbar_button_color": ["gray55", "gray41"],
+    "scrollbar_button_hover_color": ["gray40", "gray53"]
+  },
+  "CTkScrollableFrame": {
+    "label_fg_color": ["gray80", "gray21"]
+  },
+  "DropdownMenu": {
+    "fg_color": ["gray90", "gray20"],
+    "hover_color": ["gray75", "gray28"],
+    "text_color": ["gray14", "gray84"]
+  },
+  "CTkFont": {
+    "macOS": {
+      "family": "Avenir",
+      "size": 12,
+      "weight": "normal"
+    },
+    "Windows": {
+      "family": "Corbel",
+      "size": 12,
+      "weight": "normal"
+    },
+    "Linux": {
+      "family": "Montserrat",
+      "size": 12,
+      "weight": "normal"
+    }
+  },
+  "RoopDropArea": {
+    "fg_color": ["gray90", "gray13"]
+  },
+  "RoopDonate": {
+    "text_color": ["#3a7ebf", "gray60"]
+  }
+}

roop/ui.py

@@ -0,0 +1,285 @@
+import os
+import sys
+import webbrowser
+import customtkinter as ctk
+from tkinterdnd2 import TkinterDnD, DND_ALL
+from typing import Any, Callable, Tuple, Optional
+import cv2
+from PIL import Image, ImageOps
+
+import roop.globals
+import roop.metadata
+from roop.face_analyser import get_one_face
+from roop.capturer import get_video_frame, get_video_frame_total
+from roop.face_reference import get_face_reference, set_face_reference, clear_face_reference
+from roop.predictor import predict_frame, clear_predictor
+from roop.processors.frame.core import get_frame_processors_modules
+from roop.utilities import is_image, is_video, resolve_relative_path
+
+ROOT = None
+ROOT_HEIGHT = 700
+ROOT_WIDTH = 600
+
+PREVIEW = None
+PREVIEW_MAX_HEIGHT = 700
+PREVIEW_MAX_WIDTH = 1200
+
+RECENT_DIRECTORY_SOURCE = None
+RECENT_DIRECTORY_TARGET = None
+RECENT_DIRECTORY_OUTPUT = None
+
+preview_label = None
+preview_slider = None
+source_label = None
+target_label = None
+status_label = None
+
+
+# todo: remove by native support -> https://github.com/TomSchimansky/CustomTkinter/issues/934
+class CTk(ctk.CTk, TkinterDnD.DnDWrapper):
+    def __init__(self, *args: Any, **kwargs: Any) -> None:
+        super().__init__(*args, **kwargs)
+        self.TkdndVersion = TkinterDnD._require(self)
+
+
+def init(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.CTk:
+    global ROOT, PREVIEW
+
+    ROOT = create_root(start, destroy)
+    PREVIEW = create_preview(ROOT)
+
+    return ROOT
+
+
+def create_root(start: Callable[[], None], destroy: Callable[[], None]) -> ctk.CTk:
+    global source_label, target_label, status_label
+
+    ctk.deactivate_automatic_dpi_awareness()
+    ctk.set_appearance_mode('system')
+    ctk.set_default_color_theme(resolve_relative_path('ui.json'))
+
+    root = CTk()
+    root.minsize(ROOT_WIDTH, ROOT_HEIGHT)
+    root.title(f'{roop.metadata.name} {roop.metadata.version}')
+    root.configure()
+    root.protocol('WM_DELETE_WINDOW', lambda: destroy())
+
+    source_label = ctk.CTkLabel(root, text=None, fg_color=ctk.ThemeManager.theme.get('RoopDropArea').get('fg_color'))
+    source_label.place(relx=0.1, rely=0.1, relwidth=0.3, relheight=0.25)
+    source_label.drop_target_register(DND_ALL)
+    source_label.dnd_bind('<<Drop>>', lambda event: select_source_path(event.data))
+    if roop.globals.source_path:
+        select_source_path(roop.globals.source_path)
+
+    target_label = ctk.CTkLabel(root, text=None, fg_color=ctk.ThemeManager.theme.get('RoopDropArea').get('fg_color'))
+    target_label.place(relx=0.6, rely=0.1, relwidth=0.3, relheight=0.25)
+    target_label.drop_target_register(DND_ALL)
+    target_label.dnd_bind('<<Drop>>', lambda event: select_target_path(event.data))
+    if roop.globals.target_path:
+        select_target_path(roop.globals.target_path)
+
+    source_button = ctk.CTkButton(root, text='Select a face', cursor='hand2', command=lambda: select_source_path())
+    source_button.place(relx=0.1, rely=0.4, relwidth=0.3, relheight=0.1)
+
+    target_button = ctk.CTkButton(root, text='Select a target', cursor='hand2', command=lambda: select_target_path())
+    target_button.place(relx=0.6, rely=0.4, relwidth=0.3, relheight=0.1)
+
+    keep_fps_value = ctk.BooleanVar(value=roop.globals.keep_fps)
+    keep_fps_checkbox = ctk.CTkSwitch(root, text='Keep target fps', variable=keep_fps_value, cursor='hand2', command=lambda: setattr(roop.globals, 'keep_fps', not roop.globals.keep_fps))
+    keep_fps_checkbox.place(relx=0.1, rely=0.6)
+
+    keep_frames_value = ctk.BooleanVar(value=roop.globals.keep_frames)
+    keep_frames_switch = ctk.CTkSwitch(root, text='Keep temporary frames', variable=keep_frames_value, cursor='hand2', command=lambda: setattr(roop.globals, 'keep_frames', keep_frames_value.get()))
+    keep_frames_switch.place(relx=0.1, rely=0.65)
+
+    skip_audio_value = ctk.BooleanVar(value=roop.globals.skip_audio)
+    skip_audio_switch = ctk.CTkSwitch(root, text='Skip target audio', variable=skip_audio_value, cursor='hand2', command=lambda: setattr(roop.globals, 'skip_audio', skip_audio_value.get()))
+    skip_audio_switch.place(relx=0.6, rely=0.6)
+
+    many_faces_value = ctk.BooleanVar(value=roop.globals.many_faces)
+    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()))
+    many_faces_switch.place(relx=0.6, rely=0.65)
+
+    start_button = ctk.CTkButton(root, text='Start', cursor='hand2', command=lambda: select_output_path(start))
+    start_button.place(relx=0.15, rely=0.75, relwidth=0.2, relheight=0.05)
+
+    stop_button = ctk.CTkButton(root, text='Destroy', cursor='hand2', command=lambda: destroy())
+    stop_button.place(relx=0.4, rely=0.75, relwidth=0.2, relheight=0.05)
+
+    preview_button = ctk.CTkButton(root, text='Preview', cursor='hand2', command=lambda: toggle_preview())
+    preview_button.place(relx=0.65, rely=0.75, relwidth=0.2, relheight=0.05)
+
+    status_label = ctk.CTkLabel(root, text=None, justify='center')
+    status_label.place(relx=0.1, rely=0.9, relwidth=0.8)
+
+    donate_label = ctk.CTkLabel(root, text='^_^ Donate to project ^_^', justify='center', cursor='hand2')
+    donate_label.place(relx=0.1, rely=0.95, relwidth=0.8)
+    donate_label.configure(text_color=ctk.ThemeManager.theme.get('RoopDonate').get('text_color'))
+    donate_label.bind('<Button>', lambda event: webbrowser.open('https://github.com/sponsors/s0md3v'))
+
+    return root
+
+
+def create_preview(parent: ctk.CTkToplevel) -> ctk.CTkToplevel:
+    global preview_label, preview_slider
+
+    preview = ctk.CTkToplevel(parent)
+    preview.withdraw()
+    preview.configure()
+    preview.protocol('WM_DELETE_WINDOW', lambda: toggle_preview())
+    preview.resizable(width=False, height=False)
+
+    preview_label = ctk.CTkLabel(preview, text=None)
+    preview_label.pack(fill='both', expand=True)
+
+    preview_slider = ctk.CTkSlider(preview, from_=0, to=0, command=lambda frame_value: update_preview(frame_value))
+
+    preview.bind('<Up>', lambda event: update_face_reference(1))
+    preview.bind('<Down>', lambda event: update_face_reference(-1))
+    return preview
+
+
+def update_status(text: str) -> None:
+    status_label.configure(text=text)
+    ROOT.update()
+
+
+def select_source_path(source_path: Optional[str] = None) -> None:
+    global RECENT_DIRECTORY_SOURCE
+
+    if PREVIEW:
+        PREVIEW.withdraw()
+    if source_path is None:
+        source_path = ctk.filedialog.askopenfilename(title='select an source image', initialdir=RECENT_DIRECTORY_SOURCE)
+    if is_image(source_path):
+        roop.globals.source_path = source_path
+        RECENT_DIRECTORY_SOURCE = os.path.dirname(roop.globals.source_path)
+        image = render_image_preview(roop.globals.source_path, (200, 200))
+        source_label.configure(image=image)
+    else:
+        roop.globals.source_path = None
+        source_label.configure(image=None)
+
+
+def select_target_path(target_path: Optional[str] = None) -> None:
+    global RECENT_DIRECTORY_TARGET
+
+    if PREVIEW:
+        PREVIEW.withdraw()
+    clear_face_reference()
+    if target_path is None:
+        target_path = ctk.filedialog.askopenfilename(title='select an target image or video', initialdir=RECENT_DIRECTORY_TARGET)
+    if is_image(target_path):
+        roop.globals.target_path = target_path
+        RECENT_DIRECTORY_TARGET = os.path.dirname(roop.globals.target_path)
+        image = render_image_preview(roop.globals.target_path, (200, 200))
+        target_label.configure(image=image)
+    elif is_video(target_path):
+        roop.globals.target_path = target_path
+        RECENT_DIRECTORY_TARGET = os.path.dirname(roop.globals.target_path)
+        video_frame = render_video_preview(target_path, (200, 200))
+        target_label.configure(image=video_frame)
+    else:
+        roop.globals.target_path = None
+        target_label.configure(image=None)
+
+
+def select_output_path(start: Callable[[], None]) -> None:
+    global RECENT_DIRECTORY_OUTPUT
+
+    if is_image(roop.globals.target_path):
+        output_path = ctk.filedialog.asksaveasfilename(title='save image output file', defaultextension='.png', initialfile='output.png', initialdir=RECENT_DIRECTORY_OUTPUT)
+    elif is_video(roop.globals.target_path):
+        output_path = ctk.filedialog.asksaveasfilename(title='save video output file', defaultextension='.mp4', initialfile='output.mp4', initialdir=RECENT_DIRECTORY_OUTPUT)
+    else:
+        output_path = None
+    if output_path:
+        roop.globals.output_path = output_path
+        RECENT_DIRECTORY_OUTPUT = os.path.dirname(roop.globals.output_path)
+        start()
+
+
+def render_image_preview(image_path: str, size: Tuple[int, int]) -> ctk.CTkImage:
+    image = Image.open(image_path)
+    if size:
+        image = ImageOps.fit(image, size, Image.LANCZOS)
+    return ctk.CTkImage(image, size=image.size)
+
+
+def render_video_preview(video_path: str, size: Tuple[int, int], frame_number: int = 0) -> ctk.CTkImage:
+    capture = cv2.VideoCapture(video_path)
+    if frame_number:
+        capture.set(cv2.CAP_PROP_POS_FRAMES, frame_number)
+    has_frame, frame = capture.read()
+    if has_frame:
+        image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+        if size:
+            image = ImageOps.fit(image, size, Image.LANCZOS)
+        return ctk.CTkImage(image, size=image.size)
+    capture.release()
+    cv2.destroyAllWindows()
+
+
+def toggle_preview() -> None:
+    if PREVIEW.state() == 'normal':
+        PREVIEW.unbind('<Right>')
+        PREVIEW.unbind('<Left>')
+        PREVIEW.withdraw()
+        clear_predictor()
+    elif roop.globals.source_path and roop.globals.target_path:
+        init_preview()
+        update_preview(roop.globals.reference_frame_number)
+        PREVIEW.deiconify()
+
+
+def init_preview() -> None:
+    PREVIEW.title('Preview [ ↕ Reference face ]')
+    if is_image(roop.globals.target_path):
+        preview_slider.pack_forget()
+    if is_video(roop.globals.target_path):
+        video_frame_total = get_video_frame_total(roop.globals.target_path)
+        if video_frame_total > 0:
+            PREVIEW.title('Preview [ ↕ Reference face ] [ ↔ Frame number ]')
+            PREVIEW.bind('<Right>', lambda event: update_frame(int(video_frame_total / 20)))
+            PREVIEW.bind('<Left>', lambda event: update_frame(int(video_frame_total / -20)))
+        preview_slider.configure(to=video_frame_total)
+        preview_slider.pack(fill='x')
+        preview_slider.set(roop.globals.reference_frame_number)
+
+
+def update_preview(frame_number: int = 0) -> None:
+    if roop.globals.source_path and roop.globals.target_path:
+        temp_frame = get_video_frame(roop.globals.target_path, frame_number)
+        if predict_frame(temp_frame):
+            sys.exit()
+        source_face = get_one_face(cv2.imread(roop.globals.source_path))
+        if not get_face_reference():
+            reference_frame = get_video_frame(roop.globals.target_path, roop.globals.reference_frame_number)
+            reference_face = get_one_face(reference_frame, roop.globals.reference_face_position)
+            set_face_reference(reference_face)
+        else:
+            reference_face = get_face_reference()
+        for frame_processor in get_frame_processors_modules(roop.globals.frame_processors):
+            temp_frame = frame_processor.process_frame(
+                source_face,
+                reference_face,
+                temp_frame
+            )
+        image = Image.fromarray(cv2.cvtColor(temp_frame, cv2.COLOR_BGR2RGB))
+        image = ImageOps.contain(image, (PREVIEW_MAX_WIDTH, PREVIEW_MAX_HEIGHT), Image.LANCZOS)
+        image = ctk.CTkImage(image, size=image.size)
+        preview_label.configure(image=image)
+
+
+def update_face_reference(steps: int) -> None:
+    clear_face_reference()
+    reference_frame_number = int(preview_slider.get())
+    roop.globals.reference_face_position += steps
+    roop.globals.reference_frame_number = reference_frame_number
+    update_preview(reference_frame_number)
+
+
+def update_frame(steps: int) -> None:
+    frame_number = preview_slider.get() + steps
+    preview_slider.set(frame_number)
+    update_preview(preview_slider.get())

roop/utilities.py

@@ -0,0 +1,149 @@
+import glob
+import mimetypes
+import os
+import platform
+import shutil
+import ssl
+import subprocess
+import urllib
+from pathlib import Path
+from typing import List, Optional
+from tqdm import tqdm
+
+import roop.globals
+
+TEMP_DIRECTORY = 'temp'
+TEMP_VIDEO_FILE = 'temp.mp4'
+
+# monkey patch ssl for mac
+if platform.system().lower() == 'darwin':
+    ssl._create_default_https_context = ssl._create_unverified_context
+
+
+def run_ffmpeg(args: List[str]) -> bool:
+    commands = ['ffmpeg', '-hide_banner', '-loglevel', roop.globals.log_level]
+    commands.extend(args)
+    try:
+        subprocess.check_output(commands, stderr=subprocess.STDOUT)
+        return True
+    except Exception:
+        pass
+    return False
+
+
+def detect_fps(target_path: str) -> float:
+    command = ['ffprobe', '-v', 'error', '-select_streams', 'v:0', '-show_entries', 'stream=r_frame_rate', '-of', 'default=noprint_wrappers=1:nokey=1', target_path]
+    output = subprocess.check_output(command).decode().strip().split('/')
+    try:
+        numerator, denominator = map(int, output)
+        return numerator / denominator
+    except Exception:
+        pass
+    return 30
+
+
+def extract_frames(target_path: str, fps: float = 30) -> bool:
+    temp_directory_path = get_temp_directory_path(target_path)
+    temp_frame_quality = roop.globals.temp_frame_quality * 31 // 100
+    return run_ffmpeg(['-hwaccel', 'auto', '-i', target_path, '-q:v', str(temp_frame_quality), '-pix_fmt', 'rgb24', '-vf', 'fps=' + str(fps), os.path.join(temp_directory_path, '%04d.' + roop.globals.temp_frame_format)])
+
+
+def create_video(target_path: str, fps: float = 30) -> bool:
+    temp_output_path = get_temp_output_path(target_path)
+    temp_directory_path = get_temp_directory_path(target_path)
+    output_video_quality = (roop.globals.output_video_quality + 1) * 51 // 100
+    commands = ['-hwaccel', 'auto', '-r', str(fps), '-i', os.path.join(temp_directory_path, '%04d.' + roop.globals.temp_frame_format), '-c:v', roop.globals.output_video_encoder]
+    if roop.globals.output_video_encoder in ['libx264', 'libx265', 'libvpx']:
+        commands.extend(['-crf', str(output_video_quality)])
+    if roop.globals.output_video_encoder in ['h264_nvenc', 'hevc_nvenc']:
+        commands.extend(['-cq', str(output_video_quality)])
+    commands.extend(['-pix_fmt', 'yuv420p', '-vf', 'colorspace=bt709:iall=bt601-6-625:fast=1', '-y', temp_output_path])
+    return run_ffmpeg(commands)
+
+
+def restore_audio(target_path: str, output_path: str) -> None:
+    temp_output_path = get_temp_output_path(target_path)
+    done = run_ffmpeg(['-i', temp_output_path, '-i', target_path, '-c:v', 'copy', '-map', '0:v:0', '-map', '1:a:0', '-y', output_path])
+    if not done:
+        move_temp(target_path, output_path)
+
+
+def get_temp_frame_paths(target_path: str) -> List[str]:
+    temp_directory_path = get_temp_directory_path(target_path)
+    return glob.glob((os.path.join(glob.escape(temp_directory_path), '*.' + roop.globals.temp_frame_format)))
+
+
+def get_temp_directory_path(target_path: str) -> str:
+    target_name, _ = os.path.splitext(os.path.basename(target_path))
+    target_directory_path = os.path.dirname(target_path)
+    return os.path.join(target_directory_path, TEMP_DIRECTORY, target_name)
+
+
+def get_temp_output_path(target_path: str) -> str:
+    temp_directory_path = get_temp_directory_path(target_path)
+    return os.path.join(temp_directory_path, TEMP_VIDEO_FILE)
+
+
+def normalize_output_path(source_path: str, target_path: str, output_path: str) -> Optional[str]:
+    if source_path and target_path and output_path:
+        source_name, _ = os.path.splitext(os.path.basename(source_path))
+        target_name, target_extension = os.path.splitext(os.path.basename(target_path))
+        if os.path.isdir(output_path):
+            return os.path.join(output_path, source_name + '-' + target_name + target_extension)
+    return output_path
+
+
+def create_temp(target_path: str) -> None:
+    temp_directory_path = get_temp_directory_path(target_path)
+    Path(temp_directory_path).mkdir(parents=True, exist_ok=True)
+
+
+def move_temp(target_path: str, output_path: str) -> None:
+    temp_output_path = get_temp_output_path(target_path)
+    if os.path.isfile(temp_output_path):
+        if os.path.isfile(output_path):
+            os.remove(output_path)
+        shutil.move(temp_output_path, output_path)
+
+
+def clean_temp(target_path: str) -> None:
+    temp_directory_path = get_temp_directory_path(target_path)
+    parent_directory_path = os.path.dirname(temp_directory_path)
+    if not roop.globals.keep_frames and os.path.isdir(temp_directory_path):
+        shutil.rmtree(temp_directory_path)
+    if os.path.exists(parent_directory_path) and not os.listdir(parent_directory_path):
+        os.rmdir(parent_directory_path)
+
+
+def has_image_extension(image_path: str) -> bool:
+    return image_path.lower().endswith(('png', 'jpg', 'jpeg', 'webp'))
+
+
+def is_image(image_path: str) -> bool:
+    if image_path and os.path.isfile(image_path):
+        mimetype, _ = mimetypes.guess_type(image_path)
+        return bool(mimetype and mimetype.startswith('image/'))
+    return False
+
+
+def is_video(video_path: str) -> bool:
+    if video_path and os.path.isfile(video_path):
+        mimetype, _ = mimetypes.guess_type(video_path)
+        return bool(mimetype and mimetype.startswith('video/'))
+    return False
+
+
+def conditional_download(download_directory_path: str, urls: List[str]) -> None:
+    if not os.path.exists(download_directory_path):
+        os.makedirs(download_directory_path)
+    for url in urls:
+        download_file_path = os.path.join(download_directory_path, os.path.basename(url))
+        if not os.path.exists(download_file_path):
+            request = urllib.request.urlopen(url)  # type: ignore[attr-defined]
+            total = int(request.headers.get('Content-Length', 0))
+            with tqdm(total=total, desc='Downloading', unit='B', unit_scale=True, unit_divisor=1024) as progress:
+                urllib.request.urlretrieve(url, download_file_path, reporthook=lambda count, block_size, total_size: progress.update(block_size))  # type: ignore[attr-defined]
+
+
+def resolve_relative_path(path: str) -> str:
+    return os.path.abspath(os.path.join(os.path.dirname(__file__), path))