Input from local LLM or hand written text with integrated Piper model for TTS real-time wav generation even on super slow CPU's

[sarutobiumon]

It would be super cool as the next gen version to integrate a super tiny local LLM that can run even on super slow CPU's as well as the world's fastest and only real-time TTS model Piper that runs smoothly on super slow CPU's in real time to generate wav files on the fly either from LLM generated text or from user input text.

Thanks and keep up the good work! @anothermartz