app.py

import tempfile ,os
import gradio as gr
from transformers import VitsModel, AutoTokenizer,pipeline
import torch
import numpy as np
import torchaudio

model = VitsModel.from_pretrained("SeyedAli/Persian-Speech-synthesis")
tokenizer = AutoTokenizer.from_pretrained("SeyedAli/Persian-Speech-synthesis")

def TTS(text):
   inputs = tokenizer(text, return_tensors="pt")
   with torch.no_grad():
        output = model(**inputs).waveform 
   with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
        torchaudio.save(fp, output, model.config.sampling_rate)
        return fp.name
iface = gr.Interface(fn=TTS, inputs="text", outputs="audio")
iface.launch(share=False)