Update app.py

app.py

@@ -125,6 +125,7 @@ def inference_video(video_path, prompt):
     # pbar.set_description("generating video: ")
     for i, file in enumerate(input_frames):
         img = cv2.imread(os.path.join("demo_temp/input_frames", file))
+        img[:] = 0
         #vis = img + np.array([0, 0, 128]) * output[i][1].transpose(1, 2, 0)
         vis = img + output[i][1].transpose(1,2,0) * 255
         vis = np.clip(vis, 0, 255)