I think music composers could benefit from composing software: i.e. they might specify different "facets" of a score, such as Parson code (next note frequency higher, lower, or equal to previous note) for melodic relationships, harmonical relationships between simultaneous notes, melodical relationships between a note in a variation of an earlier bar and a corresponding note in the same position of the earlier bar (i.e. the melodic relationship between the expected and actual surprising note) etc...
I am writing a Z3 music composer right now actually. Just as a toy project but the idea is solid: you have a passage that you need to harmonize? Here are three different versions that conform to the rules of fuxian counterpoint, etc
Oh cool. Yeah, I looked around from "Music Composition Constraint Solving/SMT/Z3" stuff and didn't find anything which was surprising. I'll take a look at Rosette too, I've heard of it a couple times but never dug in
I have thought about that. The outputs from Markov chains and adversarial networks all sound like crap.
Despite 15 generations attempting to codify tonal harmony, there are only a few relatively few hard rules, followed by an unending literature of exceptions.
