ChatTTS
A local text-to-speech model for English and Chinese dialogue, with multiple speakers and delivery controls, restricted to noncommercial research.
ChatTTS generates conversational speech for developers and researchers building spoken LLM assistants or dialogue experiments. The released model is restricted to academic use and cannot be used commercially.
- English and Chinese speech: Generate dialogue in either supported language using pretrained models.
- Multiple speakers: Produce conversations with different voices taking turns.
- Delivery controls: Shape pauses, laughter and interjections as well as the spoken words.
The Python repository provides the model infrastructure and basic examples rather than a finished consumer voice app. Pretrained models are available through Hugging Face, and local inference can save generated audio as MP3 files.
The code is open source under AGPL-3.0. The model has a separate CC BY-NC 4.0 license for educational and research use. The authors deliberately limit the released model's audio quality through high-frequency noise and compression.
