There are various size options for the models. The choice of which you use trades off accuracy for higher performance. There is also a C++ re-implementation th

There are various size options for the models. The choice of which you use trades off accuracy for higher performance. There is also a C++ re-implementation that performs well and can definitely transcribe in realtime on many machines: https://github.com/ggerganov/whisper.cpp https://github.com/ggerganov/whisper.cpp