Live recording & Auto-segmentation, for midi notes triggering

Hi there,

I’m currently building an autoslicer with MIDI notes that trigger slices

Beside the whole record/play part with groove~, auto-crop, wave~ triggering slices with a cheap windowing thing, which works for one voice right now but should be pushed into rnbo~ for polyphony quite soon… I’d need some help with the auto-segmentation part.

Of course, I get that there is no ideal setting.
I auto normalize my recording, I have nice amp levels.

I’d like to implement a basic setting that can be used with voice, especially.
Slicing words… wow. Yes. Depending on the way we speak, it can work or not.
Imperfections allowed.

Wouldn’t someone here have done something like this ? and could help with settings (which algo for slicing, which threshold etc ?)

bufonsetslice~ seems nice
but what about novelty, transient, amp ?

I’d be happy to discuss about that.

Best regards,
Julien

Anyone?
Any advice would be appreciated.

Sorry, missed this!

For segmenting voice, I’d try bufnoveltyslice~ with the mfcc algorithm. Really hard to give advice on thresholds, but if you’re wanting whole words it might be that you’re better off using bufnoveltyfeature~ and playing around a bit – you might need more sophisticated thresholding than the slicer uses.

also, if you have an idea on how many slices you have in there (a range) you can use the iterative code example in the example folder. I coded something that uses javascript to find a given number of slices by tweaking the threshold iteratively…

1 Like

Hi @tremblap, just checked this out. Thanks.
Didn’t find it before.
I’ll try a path through it in my process.

Actually, the task could be hard to do considering time scheduled for that project.
We could fall back to something a bit different.

1 Like