add duration=300

webgui.py

@@ -189,7 +189,7 @@ def face_detection(uploaded_img, facemask_dilation_ratio, facecrop_dilation_rati
     print('face detect done.')
     return face_img, face_mask
 
-@spaces.GPU
+@spaces.GPU(duration=300)
 def video_pipe(face_img, face_mask, uploaded_audio, width, height, length, context_frames, context_overlap, cfg, steps, sample_rate, fps, device):
     face_mask_tensor = torch.Tensor(face_mask).to(dtype=weight_dtype, device="cuda").unsqueeze(0).unsqueeze(0).unsqueeze(0) / 255.0
     ref_image_pil = Image.fromarray(face_img[:, :, [2, 1, 0]])