Stuff to do Only save the sample, not the whole file Do fft on all sound samples Some actual machine learning