app.py

import streamlit as st

@st.cache_resource()
def load_models():
    from transformers import pipeline
    from TTS.api import TTS

    video_gen = pipeline('text-to-video-generation', model='cerspense/zeroscope_v2_576w')
    tts_model = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", progress_bar=False)
    return video_gen, tts_model

video_gen, tts_model = load_models()