It supports all of those.

Quite interesting thanks for sharing. Looks like a lot of work both on the implementation side and for people who want to use it, but definitely a big improvement comparing to what is default now (i.e. the robotic MIDI playing of e.g. mutopia project).

I confess that a few years ago (frustrated by how bad computer rendering of mutopia sounded) I wanted to embark myself in exactly this... too bad we did not meet back then, we might have collaborated on it. At this time in my life I am gone in the direction that you describe "your music friend's position", but I still appreciate this project and I may give it a spin once in a while.

Thanks for sharing it!

Numula is a Python library for making computer-rendered music with complex, layered variation in dynamics, timing, articulation and pedaling, as is typical in human performance.  Numula can be used to render existing compositions or for algorithmic composition.  It provides textual "shorthand" notations for scores and nuance.  The output of a Numula program is a MIDI file, which can be played with Pianoteq.

Numula is open-source and is on Github: https://github.com/davidpanderson/Numula/wiki

Can you please elaborate on how the "variation" in dynamics, timing, articulation and pedaling is achieved? Should it manually specified, falling back to a default if it's not? Does it use some randomness? Does is use some "rules" based on time signature, beat in the measure and/or measure count in the phrase?

Thanks!

Numula is open-source and is on Github: https://github.com/davidpanderson/Numula/wiki