import gradio as gr
from transformers import pipeline
import numpy as np

asr_model = "distil-whisper/distil-medium.en"

asr_pipe = pipeline("automatic-speech-recognition", model=asr_model)

def transcribe(stream, new_chunk):
    sr, y = new_chunk
    y = y.astype(np.float32)
    y /= np.max(np.abs(y))

    if stream is not None:
        stream = np.concatenate([stream, y])
    else:
        stream = y
    return stream, asr_pipe({"sampling_rate": sr, "raw": stream})["text"]

demo = gr.Blocks()


mic = gr.Interface(
    fn = transcribe,
    inputs = [
        "state", gr.Audio(sources=["microphone"], streaming=True)],
    outputs = ["state", "text"],
    layout="horizontal",
    theme="huggingface",
    title="Whisper & BERT demo - Intent Classification",
    description=(
        "Transcribe audio inputs with Whisper ASR model and detect intention from the text. Use BERT NLP model to classify the intention 
            "as one of the commands to command a light."
    ),
    allow_flagging="never",
    live=True,
)

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