once we get realtime voice without need for TTS and STT (in prod) that will be awesome, less multiple dependencies