app.py

!pip install pydub
import gradio as gr
import whisper
from gtts import gTTS
import soundfile as sf
import numpy as np
import io
import tempfile
import os
from pydub import AudioSegment

# Load the Whisper model
whisper_model = whisper.load_model("medium")

# Define transcription function
def transcribe_and_generate_response(audio_input):
    try:
        if audio_input is None:
            return "No audio input detected.", None

        # Convert audio to WAV using pydub
        audio = AudioSegment.from_file(audio_input)
        temp_audio_path = tempfile.NamedTemporaryFile(suffix=".wav", delete=False).name
        audio.export(temp_audio_path, format="wav")

        # Whisper transcribe
        result = whisper_model.transcribe(temp_audio_path, language="ur")
        transcription = result.get('text')

        if transcription is None:
            return "Transcription failed.", None

        # Get a reply from a chatbot model here (replace with your chatbot logic)
        chatbot_response = f"Your input was: {transcription}"  # Replace with actual chatbot response logic

        # Text-to-speech with gTTS
        response_audio = io.BytesIO()
        tts = gTTS(text=chatbot_response, lang="ur")
        tts.write_to_fp(response_audio)
        response_audio.seek(0)

        # Save audio to a temporary file
        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_audio_file:
            sf.write(temp_audio_file.name, np.frombuffer(response_audio.read(), dtype=np.int16), 22050)
            temp_audio_path_response = temp_audio_file.name

        return transcription, temp_audio_path_response

    except Exception as e:
        return f"An error occurred during processing: {str(e)}", None

# Gradio Interface
with gr.Blocks() as demo:
    audio_input = gr.Audio(type="filepath", label="Upload Audio")
    transcription_output = gr.Textbox(label="Transcription")
    chatbot_response_audio = gr.Audio(label="Chatbot Response", type="filepath")

    submit_btn = gr.Button("Submit")
    submit_btn.click(transcribe_and_generate_response, inputs=audio_input, outputs=[transcription_output, chatbot_response_audio])

demo.launch()