app.py

import gradio as gr
import whisper
import torch
from transformers import pipeline
from pydub import AudioSegment

#Set the device
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

# Load Whisper model
model = whisper.load_model("small", device=DEVICE)

# Load summarization pipeline
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")

def process_audio(audio_file):
    # Convert audio to WAV format if not already
    audio = AudioSegment.from_file(audio_file)
    audio = audio.set_channels(1).set_frame_rate(16000)
    temp_wav_path = "temp.wav"
    audio.export(temp_wav_path, format="wav")
    
    # Transcribe the audio
    result = model.transcribe(temp_wav_path)
    transcript = result["text"]
    
    # Summarize the transcript
    summary = summarizer(transcript, max_length=100, min_length=25, do_sample=False)[0]["summary_text"]
    return transcript, summary

# Gradio Interface
interface = gr.Interface(
    fn=process_audio,
    inputs=gr.Audio(type="filepath"),
    outputs=[
        gr.Textbox(label="Transcript"),
        gr.Textbox(label="Summary")
    ],
    title="Audio Summarization App",
    description="Upload an audio file to get its transcription and summary."
)

if __name__ == "__main__":
    interface.launch()