Update module

app.py

@@ -94,13 +94,12 @@ def run_avse_inference(video_path, audio_path):
         cv2.cvtColor(frames[i], cv2.COLOR_RGB2GRAY) for i in range(len(frames))
     ]).astype(np.float32)
     bg_frames /= 255.0
-    bg_frames = np.expand_dims(bg_frames, axis=0)
 
 
     # Combine into input dict (match what model.enhance expects)
     data = {
         "noisy_audio": noisy,
-        "video_frames": bg_frames
+        "video_frames": bg_frames[np.newaxis, ...]
     }
 
     with torch.no_grad():