Segmenting according different criteria and replaying it usind the 2D

Hi dear flucoma,

I’d like to

  • load a sample
  • segment it (I mean, finding time start/end points of parts of it) according to several criteria and then, map it on te 2D map (the jsui object based on fluid.plotter.js if I remember correctly)
  • probably segment it with different criteria at the same time (from one sample, multiple 2D maps)
  • store the analysis somewhere in order to retrieve it when I reload the sample without having to re-analyze it
  • play some segments

For playing multiple segments at the same time, using some polyphony here would be something I can understand, I think : using a poly~ with a wave~ inside of it, able to read a segment, could be easy to do.

My Main question here:

how to choose and trigger the analysis according some criteria ?

Of course, I can also understand it raises another question: how to choose the criteria for this or that sample ? Maybe, some samples could contains very “distant” segment, loudness wise. I mean, some segment would be VERY LOUD, some other very quiet. But for some others, loudness would just be… almost the same around a mean.

This question could be solved, by pre-analysing samples before choosing criteria.
It would suppose to build a small patch, for loading a sample, analysing this, that, that, that with these parameters (threshold? levels?) and check. if it seems ok, then I can trigger the analysis for this criteria and these parameters, and store that.

Then go to the other sample, and do the same. It wouldn’t have to be automatized. For this project, I don’t want to do that on A BATCH of files. Just one by one. This is totally ok in my process to do that no automated.

As far as I’d achieve this, I’d focus on splitting more than one analysis (criteria+parameters) to have multiple 2D Explorer, and to play with them with my cursor, but also automatically using some sequencers. The main work will be in that part. Using some features I saw in your patches: moving a cursor, and convoking the closest sample or even the closest sampleS…

I’ll post some other questions through more specific threads in order to make it more atomic and readable :slight_smile:

Hi Julien,

Perhaps you could run a first analysis on the whole buffer and, using the standard deviation, identify which descriptors have the greater range. These descriptors that vary widely could then be associated to the different axis of your 2D plotters.

Dominic

1 Like