app.py

import gradio as gr
import torch
from transformers import AutoProcessor, VitsModel

model_id = "facebook/mms-tts-ava"

processor = AutoProcessor.from_pretrained(model_id)
model = VitsModel.from_pretrained(model_id)

def tts_fn(text):
    inputs = processor(text=text, return_tensors="pt")
    with torch.no_grad():
        output = model(**inputs)
    waveform = output.waveform.squeeze().numpy()
    return (16000, waveform)

gr.Interface(
    fn=tts_fn,
    inputs=gr.Textbox(label="Nhập văn bản Avar"),
    outputs=gr.Audio(label="Kết quả TTS", type="numpy"),
    title="TTS Avar - facebook/mms-tts-ava"
).launch()