app.py

import gradio as gr
import torch
from transformers import pipeline

@spaces.GPU(duration=60)

# Check if GPU is available; fallback to CPU if not
# device = 0 if torch.cuda.is_available() else -1

try:
    # Load models with error handling
    transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=cuda)
    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
except Exception as e:
    print(f"Error loading models: {e}")
    raise

# Function to process audio
def process_audio(audio_file):
    try:
        # Transcribe the audio
        transcription = transcriber(audio_file)["text"]
        # Summarize the transcription
        summary = summarizer(transcription, max_length=50, min_length=10, do_sample=False)[0]["summary_text"]
        return transcription, summary
    except Exception as e:
        return f"Error processing audio: {e}", ""

# Gradio Interface with Horizontal Layout
with gr.Blocks() as interface:
    with gr.Row():
        with gr.Column():
            audio_input = gr.Audio(type="filepath", label="Upload Audio File")
            process_button = gr.Button("Process Audio")
        with gr.Column():
            transcription_output = gr.Textbox(label="Full Transcription", lines=10)
            summary_output = gr.Textbox(label="Summary", lines=5)

    process_button.click(
        process_audio,
        inputs=[audio_input],
        outputs=[transcription_output, summary_output]
    )

# Launch the interface with optional public sharing
interface.launch(share=True)