app.py

import gradio as gr
from transformers import pipeline, TFAutoModelForSeq2SeqLM, AutoTokenizer
import torch

# Initialize Hugging Face pipelines
speech_to_text = pipeline("automatic-speech-recognition", model="openai/whisper-large")
text_to_speech = pipeline("text-to-speech", model="facebook/tacotron2", device=0)  # Set device to CPU (0) or GPU (cuda)

# Function to process speech to text and text to speech
def process_audio(input_audio):
    # Convert the audio to text using Whisper model (speech-to-text)
    recognized_text = speech_to_text(input_audio)["text"]
    print(f"Recognized text: {recognized_text}")
    
    # Process the text to speech using Tacotron2 model
    audio_response = text_to_speech(recognized_text)
    return audio_response, recognized_text

# Gradio Interface for the app
def create_gradio_interface():
    with gr.Blocks() as demo:
        gr.Markdown("## AI Voice Bot for Food Ordering")

        # Audio Input: User speaks into microphone or uploads a file (filepath)
        audio_input = gr.Audio(type="filepath", label="Speak to the bot (Upload or Record Audio)")

        # Display the bot's response after recognition
        output_audio = gr.Audio(label="Bot Response", type="numpy")
        output_text = gr.Textbox(label="Bot Response (Text)")

        # Define the button to process the audio input
        audio_input.change(fn=process_audio, inputs=audio_input, outputs=[output_audio, output_text])

    return demo

# Create and launch the Gradio app
if __name__ == "__main__":
    app = create_gradio_interface()
    app.launch(share=True)