app.py

from funasr_onnx import Fsmn_vad, Paraformer, CT_Transformer
from transcribe import get_models, transcribe
import soundfile
import gradio as gr
import pytube as pt
import datetime
import os

asr_model, vad_model, punc_model = get_models("./models")

def convert_to_wav(in_filename: str) -> str:
    """Convert the input audio file to a wave file"""
    out_filename = in_filename + ".wav"
    if '.mp3' in in_filename:
        _ = os.system(f"ffmpeg -y -i '{in_filename}' -acodec pcm_s16le -ac 1 -ar 16000 '{out_filename}'")
    else:
        _ = os.system(f"ffmpeg -hide_banner -y -i '{in_filename}' -ar 16000 '{out_filename}'")
    speech, _ = soundfile.read(out_filename)
    print(f"load speech shape {speech.shape}")
    return speech

def file_transcribe(microphone, file_upload):
    warn_output = ""
    if (microphone is not None) and (file_upload is not None):
        warn_output = (
            "WARNING: You've uploaded an audio file and used the microphone. "
            "The recorded file from the microphone will be used and the uploaded audio will be discarded.\n"
        )

    elif (microphone is None) and (file_upload is None):
        return "ERROR: You have to either use the microphone or upload an audio file"

    file = microphone if microphone is not None else file_upload

    speech = convert_to_wav(file)

    items = []
    vad_model.vad_scorer.AllResetDetection()
    for item in transcribe(speech, asr_model, vad_model, punc_model):
        items.append(item)
        print(item)

    text = "\n".join(items)

    return warn_output + text


def _return_yt_html_embed(yt_url):
    video_id = yt_url.split("?v=")[-1]
    HTML_str = (
        f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"> </iframe>'
        " </center>"
    )
    return HTML_str


def youtube_transcribe(yt_url):
    yt = pt.YouTube(yt_url)
    html_embed_str = _return_yt_html_embed(yt_url)
    stream = yt.streams.filter(only_audio=True)[0]
    filename = f"audio.mp3"
    stream.download(filename=filename)

    speech=convert_to_wav(filename)
    items = []
    vad_model.vad_scorer.AllResetDetection()
    for item in transcribe(speech, asr_model, vad_model, punc_model):
        items.append(item)
        print(item)

    text = "\n".join(items)
    os.system(f"rm -rf audio.mp3 audio.mp3.wav")
    return html_embed_str, text


def run():
    gr.close_all()
    demo = gr.Blocks()

    mf_transcribe = gr.Interface(
        fn=file_transcribe,
        inputs=[
            gr.inputs.Audio(source="microphone", type="filepath", optional=True),
            gr.inputs.Audio(source="upload", type="filepath", optional=True),
        ],
        outputs="text",
        layout="horizontal",
        theme="huggingface",
        title="ParaformerX: Copilot for Audio",
        description=(
            "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the the pretrained paraformer model to transcribe audio files of arbitrary length."
        ),
        allow_flagging="never",
    )

    yt_transcribe = gr.Interface(
        fn=youtube_transcribe,
        inputs=[gr.inputs.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL")],
        outputs=["html", "text"],
        layout="horizontal",
        theme="huggingface",
        title="Demo: Transcribe YouTube",
        description=(
            "Transcribe long-form YouTube videos with the click of a button! Demo uses the the pretrained paraformer model to transcribe audio files of arbitrary length."
        ),
        allow_flagging="never",
    )

    with demo:
        gr.TabbedInterface([mf_transcribe, yt_transcribe], ["Transcribe Audio", "Transcribe YouTube"])

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

if __name__ == "__main__":
    run()