What is the behaviour of the spectral processes when there are less samples than what is asked for in the FFT or window size? I often have short short sounds so I’m curious what the behaviour is. Is it padding/stuffing to make it one frame?
good question. If I can tease you, you can try it with any object and the helpfile of bufpitch. it gives you the formula with numFrames and winSize and hopSize. Feed it a buffer of less than winSize, one less than fttSize and one less than hopSize and you’ll see how many frames you get… you can zeropad yourself and you should get the same results.
That is to say, I think your assumption is write and we should write it somewhere 