requirements

app.py

@@ -629,6 +629,13 @@ def process_video(video_input, audio_input, max_num_seconds):
         duration=10,
     )
 
+    global \
+        vae_model, \
+        hubert_model, \
+        wavlm_model, \
+        landmarks_extractor, \
+        keyframe_model, \
+        interpolation_model
     if vae_model is None:
         vae_model = VaeWrapper("video")
         vae_model = vae_model.half()  # Convert to half precision