Use VibeVoice-Large-Q8 model instead, to lower VRAM usage

app.py

@@ -48,8 +48,8 @@ try:
     logger.info("Loading VibeVoice-Large model once. This may take a while on the first run...")
     # Load the model into one node first.
     multi_speaker_node.load_model(
-        model_name='VibeVoice-Large',
-        model_path='aoi-ot/VibeVoice-Large',
+        model_name='VibeVoice-Large-Q8',
+        model_path='FabioSarracino/VibeVoice-Large-Q8',
         attention_type='auto'
     )