Update app.py

app.py

@@ -33,6 +33,7 @@ def format_classname(classname):
 
 def audio_tag(
     audio_path,
+    human_input,
     sample_rate=32000,
     window_size=800,
     hop_size=320,
@@ -59,11 +60,11 @@ def audio_tag(
     # Print audio tagging top probabilities
 
     label = labels[sorted_indexes[0]]
-    return formatted_message(format_classname(label))
+    return formatted_message(format_classname(label), human_input)
 
     
 
-def formatted_message(audio_class):
+def formatted_message(audio_class, human_input):
     if cached_audio_class != audio_class:
         cached_audio_class = audio_class
         prefix = f"""You are going to act as a magical tool that allows for humans to communicate with non-human entities like 
@@ -107,13 +108,16 @@ def formatted_message(audio_class):
 
 demo = gr.Interface(
     audio_tag,
-    gr.Audio(source="upload", type="filepath", label="Your audio"),
+    [
+        gr.Audio(source="upload", type="filepath", label="Your audio"),
+        gr.Textbox(),
+    ],
     gr.Textbox(),
     title="AnyChat",
     description="""
     <div style='text-align: center; width:100%; margin: auto;'>
-        <img src='./logo.png' alt='anychat' />
+        <img src='./logo.png' alt='anychat' width='250px' />
         <h3>Non-Human entities have many things to say, listen to them!</h3>
     </div>
-    """),
+    """,
 ).launch(debug=True)