ChatTTS

Why we included this project

ChatTTS is a text-to-speech model aimed at conversational dialogue rather than flat narration. It predicts prosodic details like laughter, interjections, and pauses, and supports multiple speakers, so a voice assistant or chatbot can sound expressive instead of robotic. Pretrained weights are included for English and Chinese, and the repo ships the core algorithm plus simple examples, which is enough to judge voice quality before you wire it into a larger LLM pipeline. Teams that want spoken output without training a voice model from scratch can start here.