Mimic 3 is an open-source neural text-to-speech engine from Mycroft AI that runs fast and entirely on local hardware. Hosted on GitHub with about 1,300 stars under the AGPL-3.0 license, it was built for Mycroft’s Mark II smart speaker and voice assistant ecosystem; the project notes it is no longer actively maintained, with the last updates in 2025.
Mimic 3 synthesizes natural-sounding speech offline, so a voice assistant never has to send text to the cloud just to speak. It ships as a Mycroft TTS plugin that can be installed with pip, or as a standalone web server distributed as a Docker image: you POST text to its local HTTP endpoint and get audio back, which makes it easy to bolt onto other projects. A separate voices repository provides a range of downloadable voices. Because everything runs locally, it works without an internet connection and keeps voice interactions private. The maintainers now point new users to Piper TTS as its spiritual successor, but Mimic 3 remains a complete, usable local TTS stack and a reference for the Mycroft ecosystem it was designed for.
Mimic 3 is free and open source under the AGPL-3.0 license. There are no per-character fees or cloud bills because synthesis happens on your own machine; you only provide the hardware it runs on.
Voice assistant builders, smart speaker tinkerers and privacy-focused developers who need dependable offline text-to-speech, especially those working within or learning from the Mycroft open-source ecosystem.