Spectral splitting

Hi all. I have several spoken voice samples (mostly male) which I wanted to split according to phonemes. Is this something that would be covered by an object such as fluid.bufonsetslice, or is there a better one? Before I try blindly, are there any specific settings that would work out better for voice?


Hi @jmmmp,

As far as the slicers go, yes, starting with the onset slicer would be a good idea, as it might respond best to the spectral changes of the different phonemes. I would probably listen to the sounds through the real-time version of the onset slicer fluid.onsetslice~ to hone in on what settings you prefer.

Phonemes can be pretty varied and very short with subtle differences, so a good exercise might be to take some of the audio file and manually slice out the phonemes, just to get a sense of the durations they tend to be and how they flow into and out of each other. That might help you hone in on what you’re looking for!



Just to elaborate on what Ted has already told you - making a patch where you can test by ear is definitely a good way to develop an intuition for how things are slicing.

quick-slice-audition.maxpat (6.3 KB)