Regarding level matching, I added this to the wiki:
We would need a standard, and perhaps two: one for polyphonic and one for monophonic instruments. For example, we could have two MIDI files (mono vs poly) where each plays a series of notes or chords, and state a goal in terms of LUFS for the resulting audio. This is scriptable (with a number of dependencies, like sfizz-render and soz.) I (Jeff Learman) will look into this. Note that this only works for sfz samplesets, but IMHO that’s what we should be targeting, using sfizz.
I’ll put that on my todo list! LMK if you disagree or have corrections/clarifications.
I’d like to see us start using “sampleset” rather than “soundfont,” especially when it’s for sfz, which is not a soundfont. I know that early on Zynthian used that term generically, but perhaps we can start cleaning that up, at least in the package info.
Or is it too late?