app.py

import gradio as gr
import nemo.collections.asr as nemo_asr
import torch

print("Starting Parakeet ASR App...")

# Use GPU if available
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

print("Downloading Parakeet model from Hugging Face...")
model = nemo_asr.models.ASRModel.from_pretrained(
    model_name="nvidia/parakeet-tdt-0.6b-v3"
)

model = model.to(device)
model.eval()

print("Model loaded successfully!")

def transcribe(audio):
    if audio is None:
        return "Please upload an audio file."

    try:
        result = model.transcribe(paths2audio_files=[audio])
        return result[0]
    except Exception as e:
        return f"Error during transcription: {str(e)}"

demo = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(type="filepath"),
    outputs="text",
    title="🎙️ Parakeet ASR (0.6B v3)",
    description="Speech-to-Text using NVIDIA Parakeet model (downloaded at runtime)"
)

demo.launch(server_name="0.0.0.0", server_port=7860)