Pin the weights' revision, and label the output as the speech it holds.

app.py

@@ -42,6 +42,7 @@ def download_weights():
         print("Downloading model weights from HuggingFace...")
         snapshot_download(
             repo_id=repo_id,
+            revision="b3a2c297eea319931ee985751f117f866085d0c5",
             local_dir=weights_dir,
             local_dir_use_symlinks=False
         )
@@ -245,7 +246,7 @@ if __name__ == "__main__":
             info="Increase this for a stronger likeness to the reference voice."
         )
 
-        output_audio = gr.Audio(type="filepath", label="Generated Audio").set_info((
+        output_audio = gr.Audio(type="filepath", label="Synthesized Speech").set_info((
             "The text spoken in the reference voice."
         ))