app.py

import gradio as gr
from transformers import pipeline
import torch
import torchaudio

device = "cuda:0" if torch.cuda.is_available() else "cpu"

def transcribe(audio):
    # Extract the audio data from the tuple
    audio_data = audio[0] if isinstance(audio, tuple) else audio
    
    # Load the audio data using torchaudio
    waveform, sample_rate = torchaudio.load(audio_data)
    
    # Convert the waveform to numpy array
    waveform_np = waveform.numpy()
    
    pipe = pipeline(
        "automatic-speech-recognition",
        model="openai/whisper-small",
        chunk_length_s=30,
        device=device,
    )
    
    prediction = pipe(waveform_np)["text"]
    return prediction

gradio_app = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(label="Input"),
    outputs=gr.Textbox(label="Result"),
    title="Transcribed",
)

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