There's another popular-ish paradigm: basically software modular synths, rather low-level so you create individual sounds by wiring the signals into graphs. Max/MSP and ‘Pure Data’ are the most famous in this style, they're often called ‘sound programming languages’ for some reason. I'm having a hard time finding a finished demonstration instead of a tutorial, but here's one for Pure Data: https://www.youtube.com/watch?v=DGiT55likdo

SunVox, mentioned elsewhere in the thread, is similar in the approach afaik, and works on many platforms (I first saw it on Palm and then Windows Mobile in mid-2000s).

CSound is sorta related somehow, but it's an actual textual language for generating sound.

There are also apps that look like hardware modular synths: VCV Rack is the most well-known and open-source; apparently Reaktor is in this vein too. And there's Hexen on Android. (Propellerhead Reason and Reaper both flirt with the modular approach, making you wire the signal between some modules, but they seem to have dialed it down, dropping unnecessary work for the user).