Fix type annotation

app.py

@@ -75,7 +75,7 @@ net, feature_utils, seq_cfg = get_model()
 @spaces.GPU(duration=120)
 @torch.inference_mode()
 def video_to_audio(
-    video: gr.Video,
+    video: str,
     prompt: str,
     negative_prompt: str,
     seed: int,
@@ -90,7 +90,7 @@ def video_to_audio(
     the MMAudio model. The output is a video file with generated audio.
 
     Args:
-        video (gr.Video): Input video used for visual and temporal conditioning.
+        video (str): Input video used for visual and temporal conditioning.
         prompt (str): Text prompt describing the desired audio content.
         negative_prompt (str): Text describing audio characteristics to avoid.
         seed (int): Random seed for reproducibility (-1 for random).