Pocket TTS
Small CPU-only text-to-speech model with voice cloning and streaming
What it is
Pocket TTS is a 100M-parameter speech synthesis model from Kyutai, with a Python API, a CLI, and a local HTTP server with a web interface. On just two CPU cores it generates speech faster than real time, streams audio with little delay, copies a voice from a short wav clip, and supports seven languages: English, French, Spanish, German, Italian, Portuguese and Dutch. Developers who want offline speech output without a GPU or a paid API can install it with pip, and the released training code lets them train their own variants.