Pure Data algorithmic composition in 53-EDO (with visualization)

This is still a work in progress, so the presentation isn't as refined as it might be. Feedback would be helpful. In particular I'd like to know what I can add, remove or change to make it easier to understand. The visualization part is described in the video's annotations.

The basic idea is to do a computational search for consonant 6 note scales, which can then be connected into scale networks (as in Dmitri Tymoczko's work). I wrote a number of Octave scripts to do this. First, all possible scales are found, then most are eliminated based on step size constraints and calculated total dissonance. The idea is to find the "useful" scales that can be made in any arbitrary equal temperament. Here I'm using 568 scales out of a total of 433160.

You'll notice that the step size isn't consistent. This isn't so great for melodic purposes, but I think it works fine here. Also, not all of the intervals are optimal. I'm trying to spread the total error between a number of less than perfect intervals rather than have a smaller number of very bad intervals (i.e. I'm using squared error). The dissonance is calculated from both the "complexity" of the ideal just interval (identified automatically, so no prior knowledge is needed), and the deviation from this in cents. This part is still sort of ad hoc, but it gives sensible results.

Most of the sounds are made by my own algorithmic additive/formant synthesis engine. I might do a video later that describes that.