OpenVoice
myshell-ai/OpenVoice · Python · MIT
Instant voice cloning with style control
OpenVoice made zero-shot cloning approachable: MIT code, simple demos, and flexible style controls. Great for prototypes; evaluate newer F5/Fish Speech for latest quality.
Score
Setup at a glance
git clone https://github.com/myshell-ai/OpenVoice && pip install -r requirements.txt Deep dive
OpenVoice is MyShell's instant voice-cloning toolkit: a short reference clip is enough to reproduce a voice, and — its signature feature — the tone colour, emotion, accent and speaking style can be controlled separately from the speaker identity.
The work comes out of MIT and Tsinghua (the author list includes Zengyi Qin at MIT and Wenliang Zhao at Tsinghua), is MIT-licensed, and has an arXiv paper documenting the approach.
It ships usage docs and a demo video, and the same tech powers voice features in the MyShell app — so what you clone locally is a research-then-product pipeline.