Training for real-time NMF (fluid.nmfmatch~)

Yeah that’s what I’m going to try next.

Mainly to see if this improves matching, and what if changed in them from @filterupdate 1, and seeing if that improves matching.

All of that, unfortunately, is dependant on the not-perfect post-processing the patch is using at the moment. So assessing the efficacy won’t actually be “possible”.

Hope the list derivative example @weefuzzy posted in the other thread will be of help (along with the newly upgraded listSchmitt).