Mimic 2 Kusal voice for self-hosting TTS service

Hi there! Shiny new Mycroft supporter here :slight_smile:

I was checking the ecosystem of TTS and STT, in general, and Mimic 2 catched my attention. However, I read something about the pre-trained Kusal voice, and how it was available for Mycroft supporters… but I haven’t been able to find it.

I would love to self-host TTS and STT servers in my home Jetson device, but I am uncertain on infrastructure for pretrained data and the configuration for server side and client side.

Have I missed some obvious documentation on that topic?

While the Mimic2 sourcecode is available, the pre-trained model for the Kusal voice itself was not published (yet?).

You can train your own voice - either from one of the public available datasets or from a recording of your own voice (which is a tedious task as you need several hours of clean recording plus text transcription).