Update app.py

app.py

@@ -60,9 +60,15 @@ def read_video(video) -> torch.Tensor:
     """
     Reads a video file and converts it into a torch.Tensor with the shape [F, C, H, W].
     """
-    
+    video_tensor = []
     to_tensor_transform = transforms.ToTensor()
-    video_tensor = torch.stack([to_tensor_transform(img) for img in video])
+    if isinstance(video, str):
+        with imageio.get_reader(video) as reader:
+            # Read all frames
+            for frame in reader:
+                video_tensor.append(to_tensor_transform(frame))
+    else: # video is a list of pil images
+        video_tensor = torch.stack([to_tensor_transform(img) for img in video])
     return video_tensor
 
 def round_to_nearest_resolution_acceptable_by_vae(height, width, vae_temporal_compression_ratio):
@@ -199,7 +205,8 @@ def generate_video(
         print("######## control_video ", control_video)
         if control_video is not None:
             # Use the pre-processed canny frames
-            processed_video = load_video(control_video)
+            #processed_video = load_video(control_video)
+            processed_video = control_video
             print("$$$$$$$ control_video is not None ", control_video)
         else:
             print("$$$$$$$ control_video is None ", reference_video, width, height)