Update webgui.py

webgui.py

@@ -155,71 +155,8 @@ def select_face(det_bboxes, probs):
     sorted_bboxes = sorted(filtered_bboxes, key=lambda x:(x[3]-x[1]) * (x[2] - x[0]), reverse=True)
     return sorted_bboxes[0]
 
-@spaces.GPU
 lmk_extractor = LMKExtractor()
-# def process_video(uploaded_img, uploaded_audio, width, height, length, seed, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device):
-
-#     if seed is not None and seed > -1:
-#         generator = torch.manual_seed(seed)
-#     else:
-#         generator = torch.manual_seed(random.randint(100, 1000000))
-
-#     #### face musk prepare
-#     face_img = cv2.imread(uploaded_img)
-#     face_mask = np.zeros((face_img.shape[0], face_img.shape[1])).astype('uint8')
-#     det_bboxes, probs = face_detector.detect(face_img)
-#     select_bbox = select_face(det_bboxes, probs)
-#     if select_bbox is None:
-#         face_mask[:, :] = 255
-#     else:
-#         xyxy = select_bbox[:4]
-#         xyxy = np.round(xyxy).astype('int')
-#         rb, re, cb, ce = xyxy[1], xyxy[3], xyxy[0], xyxy[2]
-#         r_pad = int((re - rb) * facemask_dilation_ratio)
-#         c_pad = int((ce - cb) * facemask_dilation_ratio)
-#         face_mask[rb - r_pad : re + r_pad, cb - c_pad : ce + c_pad] = 255
-        
-#         #### face crop
-#         r_pad_crop = int((re - rb) * facecrop_dilation_ratio)
-#         c_pad_crop = int((ce - cb) * facecrop_dilation_ratio)
-#         crop_rect = [max(0, cb - c_pad_crop), max(0, rb - r_pad_crop), min(ce + c_pad_crop, face_img.shape[1]), min(re + r_pad_crop, face_img.shape[0])]
-#         face_img = crop_and_pad(face_img, crop_rect)
-#         face_mask = crop_and_pad(face_mask, crop_rect)
-#         face_img = cv2.resize(face_img, (width, height))
-#         face_mask = cv2.resize(face_mask, (width, height))
-
-#     ref_image_pil = Image.fromarray(face_img[:, :, [2, 1, 0]])
-#     face_mask_tensor = torch.Tensor(face_mask).to(dtype=weight_dtype, device="cuda").unsqueeze(0).unsqueeze(0).unsqueeze(0) / 255.0
-    
-#     video = pipe(
-#         ref_image_pil,
-#         uploaded_audio,
-#         face_mask_tensor,
-#         width,
-#         height,
-#         length,
-#         steps,
-#         cfg,
-#         generator=generator,
-#         audio_sample_rate=sample_rate,
-#         context_frames=context_frames,
-#         fps=fps,
-#         context_overlap=context_overlap
-#     ).videos
-
-#     save_dir = Path("output/tmp")
-#     save_dir.mkdir(exist_ok=True, parents=True)
-#     output_video_path = save_dir / "output_video.mp4"
-#     save_videos_grid(video, str(output_video_path), n_rows=1, fps=fps)
-
-#     video_clip = VideoFileClip(str(output_video_path))
-#     audio_clip = AudioFileClip(uploaded_audio)
-#     final_output_path = save_dir / "output_video_with_audio.mp4"
-#     video_clip = video_clip.set_audio(audio_clip)
-#     video_clip.write_videofile(str(final_output_path), codec="libx264", audio_codec="aac")
-
-#     return final_output_path
-
+@spaces.GPU
 def process_video(uploaded_img, uploaded_audio, width, height, length, facemask_dilation_ratio, facecrop_dilation_ratio, context_frames, context_overlap, cfg, steps, sample_rate, fps, device):
     #### face musk prepare
     face_img = cv2.imread(uploaded_img)