change to cpu

app.py

@@ -175,8 +175,8 @@ with gr.Blocks(title="Voice chat with LLM") as demo:
                 # generate speech by cloning a voice using default setting
                 generated_speech = generate_speech_for_sentence(chatbot_history, chatbot_voice, sentence, xtts_model, None, return_as_byte=True)
                 if generated_speech is not None:
-                    _, audio_dict = generated_speech
-                    yield (sentence, chatbot_history, audio_dict)
+                    #_, audio_dict = generated_speech
+                    yield (sentence, chatbot_history, generated_speech)
                     #yield (sentence, chatbot_history, audio_dict["value"])
 
         if initial_greeting: