app.py

import torch

import gradio as gr
import pytube as pt
from transformers import pipeline
from huggingface_hub import model_info

MODEL_NAME = "cloudqi/cqi_speech_recognize_pt_v0"

device = "cuda" if torch.cuda.is_available() else "cpu"

pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    chunk_length_s=30,
    device=device,
)

langs = model_info(MODEL_NAME).cardData["language"]

article = f"<details><summary>Esse modelo suporta {len(langs)} línguas ! (Clique para expandir)</summary>> {langs}</details>"

def transcribe(microphone, file_upload):
    warn_output = ""
    if (microphone is not None) and (file_upload is not None):
        warn_output = (
            "WARNING: Você carregou um arquivo de áudio e usou o microfone. "
            "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: Transcreva microfones longos ou entradas de áudio com o clique de um botão"

    file = microphone if microphone is not None else file_upload

    text = pipe(file)["text"]

    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 yt_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]
    stream.download(filename="audio.mp3")

    text = pipe("audio.mp3")["text"]

    return html_embed_str, text


demo = gr.Blocks()

mf_transcribe = gr.Interface(
    fn=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="Demonstração: Transcrever Audio",
    description=(
        "Transcreva microfones longos ou entradas de áudio com o clique de um botão! Essa Demo usa o ajuste fino"
        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) e 🤗 Transformers para transcrever arquivos de áudio"
        " de comprimento arbitrário."
    ),
    article=article,
    allow_flagging="never",
)

yt_transcribe = gr.Interface(
    fn=yt_transcribe,
    inputs=[gr.inputs.Textbox(lines=1, placeholder="Cole o URL de um vídeo do YouTube aqui", label="YouTube URL")],
    outputs=["html", "text"],
    layout="horizontal",
    theme="huggingface",
    title="Transcrever do YouTube",
    description=(
        "Gere legendas com um clique ! A demonstração usa o ponto de verificação aprimorado:"
        f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) e 🤗 Transformers para transcrever arquivos de áudio de"
        " comprimento arbitrário."
    ),
    article=article,
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface([mf_transcribe, yt_transcribe], ["Transcrever de áudio", "Transcrever do YouTube"])

demo.launch(enable_queue=True)