Update app.py

app.py

@@ -126,7 +126,13 @@ def model(text, web_search):
         return "".join([response.token.text for response in stream if response.token.text != "</s>"])
 
 async def respond(audio, web_search):
-    user = transcribe(audio)
+    audio_data, sr = audio  # Unpack the audio tuple
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
+        tmp_path = tmp_file.name
+        audio_segment = AudioSegment(data=audio_data, sample_width=2, frame_rate=sr, channels=1)  # Assuming 16-bit PCM audio
+        audio_segment.export(tmp_path, format="wav")
+
+    user = transcribe(tmp_path)
     reply = model(user, web_search)
     communicate = edge_tts.Communicate(reply)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
@@ -139,7 +145,7 @@ with gr.Blocks() as demo:
         web_search = gr.Checkbox(label="Web Search", value=False)
         input = gr.Audio(label="Voice Chat", sources="microphone")
         output = gr.Audio(label="AI",autoplay=True)
-        gr.Interface(fn=respond, inputs=[input, web_search], outputs=[output], live=True, batch=True, max_batch_size=20, delete_cache=(60,60))
+        gr.Interface(fn=respond, inputs=[input, web_search], outputs=[output], live=True)
  
 if __name__ == "__main__":
     demo.queue(max_size=200).launch()