app.py

import gradio as gr
import nemo.collections.asr as nemo_asr
from nemo.collections.asr.models import ASRModel
import os
from huggingface_hub import hf_hub_download

# Download model from private repo
model_path = hf_hub_download(
    repo_id="KnowlAI/parakeet-rnnt-0.6b-08122025-11.55",
    filename="parakeet-knowl-hindi.nemo",
    token=os.environ.get("HF_TOKEN")
)
asr_model = ASRModel.restore_from(model_path)

def transcribe_audio(audio):
    """Transcribe audio file"""
    if audio is None:
        return "Please upload or record an audio file."

    try:
        # Transcribe
        transcription = asr_model.transcribe([audio])
        return transcription[0]
    except Exception as e:
        return f"Error: {str(e)}"

def greet(name):
    return "Hello " + name + "!!"

demo = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(type="filepath", label="Upload or Record Audio"),
    outputs=gr.Textbox(label="Transcription (Hindi)", lines=5),
    title="Parakeet Hindi ASR Demo",
    description="Upload or record Hindi speech to get the transcription. This model was trained for 100,000 steps on Hindi speech data."
)

#demo = gr.Interface(fn=greet, inputs="text", outputs="text")
demo.launch()