app.py

import gradio as gr
import re
import subprocess
import math
import shutil
import soundfile as sf
import tempfile
import os
import requests
import time


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 transcribe_base(audio, language):
    start_time = time.time()

    d, sr = sf.read(audio)
    data = {'audio': d.tolist(),
            'sampling_rate': sr,
            'language': language}

    response = requests.post(os.getenv("api_url"), json=data).json()
    transcription = response["text"]
    speaker_class_string = response["speaker_class_string"]

    end_time = time.time()
    print("-"*50)
    print(len(data["audio"])/float(sr))
    print(end_time-start_time)
    print("-"*50)

    return transcription, speaker_class_string


def transcribe(audio_microphone, audio_upload, language):
    print("Transcription request")
    print(audio_microphone, audio_upload, language)
    audio = audio_microphone if audio_microphone is not None else audio_upload
    return transcribe_base(audio, language)


demo = gr.Blocks()
with demo:
    gr.Markdown("# Speech recognition using Whisper models")
    gr.Markdown("Orai NLP Technologies")

    with gr.Tab("Trancribe Audio"):
        iface = gr.Interface(
            fn=transcribe,
            inputs=[
                gr.Audio(sources="microphone", type="filepath"),
                gr.Audio(sources="upload", type="filepath"),
                gr.Dropdown(choices=[("Basque", "eu"),
                                     ("Spanish", "es"),
                                     ("English", "en")],
                                     #("French", "fr"),
                                     #("Italian", "it"),
                            value="eu")
            ],
            outputs=[
                gr.Textbox(label="Transcription", autoscroll=False),
                gr.Textbox(label="Speaker Identification", autoscroll=False)
            ],
            allow_flagging="never",
        )
demo.queue(max_size=1)
demo.launch(share=False, max_threads=3, auth=(os.getenv("username"), os.getenv("password")), auth_message="Please provide a username and a password.")