When I listen to the great results from FluidSines, I keep thinking we could also have resonance analysis.
We already super briefly talked about resonance models. I do not know if this is relevant here but that would be another mean of analysis/classification/resynthesis using fof, Loris’ oscillator, reson filters etc. That is strait from the 1980’s but transient synthesis can then sound good. Just a starting point, or just fantasy.
fluid.bufnmf~ makes me think of iterative techniques from a modres analysis. That may be some connections here. As a start, fluid.bufnmf~, with a time factor, may also become some sort of modres analysis ?
The question would be how and where to store and access them in Max: index, freq, amp, bandwidth. dict, signal ?
Hi, this is indeed an interesting avenue, although I don’t think we are going to be able to explore it short term. There are plans for some improvements to FluidSines, mainly for better estimation. But thanks for the ideas, we’ll keep them in mind! Another interesting sort of related avenue is source-filter model formulations of NMF (and NTF)…
I know the waveshaping could be improved (I just tweaked/adapted the waveshaper from BEAP so it’s much cheaper), and probably some better overall structuring to make it cost less overall, but it is an expensive process.
maxmodres Max Package:
A resonance model analysis. Jean Lochard, made a Max patch to make the shell commands for a modres-2.0.0 running for sure under Mac OS 10.13 and more: https://mega.nz/#!wvRgkSiY!Kpep2NJyfgGDxUQcjUxsBibBTZgYSOd1mvHg5un05vE
The modres binary is within the zip file. It is officially freely available online.