Comment by steinvakt2

Any way to use this for speech-to-text? To gain faster whisper inference for instance, without losing accuracy?