app.py

from transformers import pipeline
import gradio as gr

# Create a pipeline for text-to-speech
tts = pipeline("text-to-speech", model="facebook/mms-tts-eng")

# Create a pipeline for speech-to-text
stt = pipeline("automatic-speech-recognition", model="openai/whisper-medium")

# Create a pipeline for text generation
chat = pipeline("text-generation", model="openai-community/gpt2-xl")

def voice_chat(user_voice):
    user_text = stt(user_voice)["text"]
    messages = [{"role": "system", "content": "You are a kind helpful assistant."}]
    messages.append({"role": "user", "content": user_text})
    chat_reply = chat(messages=messages, max_length=100, top_p=0.95, temperature=0.7)[0]["generated_text"]
    messages.append({"role": "assistant", "content": chat_reply})
    audio = tts(chat_reply)["audio"]
    return chat_reply, audio

text_reply = gr.Textbox(label="ChatGPT Text") 
voice_reply = gr.Audio('output.wav')

gr.Interface(
    title="AI Voice Assistant with ChatGPT AI",
    fn=voice_chat,
    inputs=[Interface.Audio(source="microphone", type="filepath")],
    outputs=[text_reply, voice_reply],
    live=True,
).launch(debug=True)