app.py

import gradio as gr
import torch
from faster_whisper import WhisperModel

# Determine compute device and model size
device = "cuda" if torch.cuda.is_available() else "cpu"
compute_type = "float16" if torch.cuda.is_available() else "float32"

# Initialize the faster-whisper model
model = WhisperModel("tiny",
    device=device,
    compute_type=compute_type
)

def transcribe(audio):
    # Transcribe audio using faster-whisper
    segments, _ = model.transcribe(audio, language="yo")
    # Combine all segments into one text
    result = " ".join([segment.text for segment in segments])
    return result

iface = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(sources=["microphone", "upload"], type="filepath"),
    outputs="text",
    live=True,
    title="Speech-to-Text Demo",
    description="Transcribe speech to text using the Whisper model."
)

if __name__ == "__main__":
    iface.launch(share=True)