Retrieve audio from inputs

app.py

@@ -52,7 +52,8 @@ def populate_metadata(link):
     audio = get_audio(link)
     return yt.thumbnail_url, yt.title, audio
 
-def inference(audio):
+def inference(inputs):
+    audio = inputs[0]
     with open(audio, "rb") as fd:
         contents = fd.read()
 
@@ -91,7 +92,7 @@ with demo:
         The mt3 model transcribes multiple simultaneous instruments, but without velocities.
         """
         model = gr.Radio(
-            ["mt3", "ismir2021"], 
+            ["mt3",], 
             label=model_label, 
             value="mt3"
         )