Chatterbox TTS

Chatterbox TTS is the MIT-licensed, open-source Text-to-Speech model from Resemble AI, delivering sub-200ms latency, zero-shot voice cloning, and unique emotion exaggeration control.

This is a production-grade, open-source TTS solution, built by Resemble AI and released under an MIT license. It stands out with a unique emotion exaggeration control parameter, allowing users to fine-tune expression from monotone to dramatically expressive. The model is engineered for real-time applications, achieving ultra-low inference latency of under 200 milliseconds. Trained on over 500K hours of data, Chatterbox consistently outperforms proprietary models in blind tests and features responsible AI safeguards: every generated audio file includes an imperceptible PerTh neural watermark for provenance tracking.

https://chatterbox.run