+1 for voxtype with Whisper-base model it is quite fast an accurate