app.py

# -*- coding: utf-8 -*-
import re
from ttsmms import TTS
import gradio as gr

tts = TTS("khm")

def sanitize(text):
    return re.sub(r"\u200b", "", text)

def generate_voice(text):
    audio = tts.synthesis(text)
    return (audio['sampling_rate'], audio['x'])

with gr.Blocks(title="Khmer Text to Speech with MMS") as blocks:

    gr.Markdown('# Khmer Text to Speech - MMS')
    gr.Markdown('MMS: Scaling Speech Technology to 1000+ languages by Meta AI')

    input_text = gr.Textbox(label="អក្សរ", lines=3)
    examples = gr.Examples(examples=["នេះ​ជា​រូបថត​ជនសង្ស័យ​ដែល​បេីកឡាន។"], inputs=[input_text])

    run_button = gr.Button(
        text="Generate", 
        type="button",
    )

    out_audio = gr.Audio(
        label="សំឡេងដែលបានបង្កើត",
        type="numpy", 
    )

    inputs = [input_text]
    outputs = [out_audio]

    run_button.click(
        fn=generate_voice,
        inputs=inputs,
        outputs=outputs,
        queue=True,
    )


blocks.queue(concurrency_count=1).launch(debug=True)