i loved working on it, really - but hebrew TTS isnt developed enough for me to continue working on this right now.
the project was ment to be my personal assistant that runs 24/7. i dont like the "Amazon Alexa"-like personal assitant who use wake words, i wanted something that listens to me 24/7 and that could detect when im speaking to IT. so, thats victor.
its is not even close to being done, hebrew TTS (at least real-time TTS) isnt ready yet for a project like this. but theres a pretty good transcription system that im really proud of that is basically ready!
the real-time transcription system victor uses is basically done, i just need to flesh it out a bit - it uses faster-whisper that transcribes small chunks of audio in real time, and uses (attempts to at least, havent figured this part out yet) realistic End-Of-Turn detection using a smaller model for realistic conversations.