Allright so this log is about research and MVP (minimal viable product)
So I stared playing with onyx-om python library and there are also some examples for python. I’ve played with them a bit and it seems that whisper is the king for STT and as for TTS I’m going for piper.

I have modified the STT code from simpler and less acurate model for whisper which is way better and I don’t mind the lag.
And after that I’ve played with different things like other models and TTS.
After that I managed to get implemented this workflow. You press t it starts recording, after release it transcribes it, and with tts piper it converts back to speach. This is meant to prove me the thesis.

So the main core is mostly done, although the code is very messy.