neonbjb/tortoise-tts ? reverse-engineered prompt

Build me a local text to speech app using this project that can turn typed text into natural sounding speech with multiple voices. I want a simple way to paste in text, choose a voice or use my own sample clips, then generate and save the audio file. It should also handle longer text by reading it sentence by sentence and combining the result, with an option to regenerate any bad parts. If possible, make it easy to run from the notebook or a simple command line flow, and include a basic streaming or live playback option for quicker testing. Keep it focused on quality and make the setup clear for someone using an NVIDIA GPU. If you need anything unclear, look up the current docs online and use the existing examples in the repo as a guide.