Generating and displaying many structs can be very slow, it might not be the most efficient way. Not sure internally if it's better to bang~ a whole table, or find a method to go not so fast. Anyway, when generating structs, it will help if the window where they will be drawn is closed.

I think the eargram library had already a sonogram also based on data structures, probably using timbreId.

I also did some own improvements to the partial tracer file (for both analysis and synthesis), there were some small bugs/incoherences there: (not sure if this picture will show in the list)

I wanted to implement a sonogram analysis like max's spectroscope~ does (see picture) with FFT analysis and data structures. Any ideas? I guess I can [fft~] connected to a [tabsend~] and then have a [bang~] reading the table values for generating the graph, that's my first guess.

 But I am also now inspired by partialtracer.pd from the data structures tutorial, which uses [sigmund~] and it'd be great if it had a new output mode for raw FFT amplitudes, I am assuming that the 'peaks' output doesn't really work for this, or would it?