app.py

import streamlit as st
from transformers import VitsModel, AutoTokenizer
import torch

# Title and Description
st.title("Text-to-Speech with VitsModel")
st.write("Enter some English text, and I'll generate audio for you!")

# Load Model and Tokenizer
@st.cache_resource  # Cache the model for efficiency
def load_tts_model():
    model = VitsModel.from_pretrained("facebook/mms-tts-eng")
    tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-eng")
    return model, tokenizer

model, tokenizer = load_tts_model()

# User Input
user_text = st.text_input("Enter your text here:")

# Generate Audio on Button Click
if st.button("Generate Speech"):
    if not user_text:
        st.warning("Please enter some text.")
    else:
        inputs = tokenizer(user_text, return_tensors="pt")
        with torch.no_grad():
            output = model(**inputs).waveform

        # Play the Audio Directly
        st.audio(output[0].numpy())