app.py

import gradio as gr
import requests
from typing import Union
from pydub import AudioSegment
from io import BytesIO
import os
from utils import CheckLanguageIsSpanish

title = """
<div style="display: flex; align-items: center;">
    <img src="https://i.postimg.cc/NGVZL3bg/logo-sintetica.png" alt="sintetica_logo" width="40" height="40" style="margin-right: 10px;">
    <h1 style="font-size: 24px;">Sintética: Text-to-Speech (TTS)</h1>
</div>
"""

description = """
<div style="font-size: 16px;">

Sintética: Sistema text-to-speech (TTS) desarrollado por MecanTronic S.A.

Para más información:
</div>

<div style="display: flex; flex-direction: column;">
    <a href="https://sintetica.com.ar/" style="display: flex; align-items: center; margin-bottom: 5px; font-size:16px;">
        <img src="https://i.postimg.cc/NGVZL3bg/logo-sintetica.png" alt="sintetica_logo" width="20" height="20" style="margin-right: 10px">
        Sintética
    </a>  
    <a href="https://mecantronic.com.ar/" style="display: flex; align-items: center; font-size:16px;">
        <img src="https://i.postimg.cc/X7Cm5sD9/logo-MEC.png" alt="mecantronic_logo" width="20" height="20" style="margin-right: 10px;">
        MecanTronic
    </a>
</div>
"""
speaker_mapping = {
    "Micaela": "arf_00295",
    "Florencia": "arf_02121",
    "Rocío": "arf_02484",
    "Pedro": "arm_03397",
    "Pablo": "arm_06136",
    "Juan": "arm_05223"}

def process( text: str, speaker: str, speed: Union[int, float]):

    speaker = speaker_mapping[speaker]

    if check_errors(text):
        return None

    request = {
        "text": text,
        "speaker": speaker,
        "pitch_scale_factor": 0.95,
        "length_scale_factor": 1/speed,
        "output_samplerate": 44100}

    audio_path = post_request(request)    
    audio_widget = play_audio(audio_path)

    return audio_widget


def post_request(request: dict):
    url = os.environ.get("URL_INFERENCE")
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=request, headers=headers)
    return BytesIO(response.content)


def play_audio(audio_bytes: BytesIO):
    audio = AudioSegment.from_file(audio_bytes)
    audio_widget = gr.Audio(audio.export(format="wav").read())
    return audio_widget


def check_errors(text):
    error = False
    if len(text) > 330:
        gr.Warning(f'La longitud del texto ({len(text)} caracteres) sobrepasa el máximo permitido.')
        error = True
    if not CheckLanguageIsSpanish().detect_english(text):
        gr.Warning('El texto está en Ingles o posee oraciones en dicho idioma.')
        error = True
    return error

        
demo = gr.Blocks()
with demo:
    gr.HTML(title)

    language_choices = ["Español"]
    language = gr.Radio(
        choices=language_choices,
        value=language_choices[0],
        label="Idioma",
        info="Próximamente nuevos idiomas disponibles"
    )

    with gr.Tabs():
        with gr.TabItem("Opciones del TTS"):
            input_text = gr.Textbox(
                label="330 caracteres como máximo",
                type="text",
                lines=3,
                placeholder="Ingresa aquí el texto"
            )

            input_speaker = gr.Dropdown(
                label="Hablante",
                choices=speaker_mapping.keys(),
                type="value",
                value=list(speaker_mapping.keys())[0]
            )

            input_speed = gr.Slider(
                minimum=0.5,
                maximum=1.5,
                value=1,
                step=0.05,
                label="Velocidad de reproducción",
            )

            with gr.Row():
                input_button = gr.Button("Procesar", variant="primary")
                clear_button = gr.ClearButton([input_text], value="Limpiar")

        gr.HTML('<div style="height: 15px;"></div>')

        output_audio = gr.Audio(
            type="filepath",
            label="Audio sintetizado"
        )

        input_button.click(
            process,
            inputs=[
                input_text,
                input_speaker,
                input_speed,
            ],
            outputs=[
                output_audio,
            ],
        )

    gr.Markdown(description)


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