For me with a small (but very accurate) whisper model it can be even faster than AquaVoice, and works locally (which is really important)