sound generation on the fly

I'm looking to generate sound on the f开发者_开发百科ly given a set of parameters, and produce something along the lines of a midi to be played as it is created. Is this the best way to go about this? The important feature I'm looking for is control over the sound of the music, i.e., the instrument. Does anyone have suggestions on good sound libraries, apis?

Note, I'd rather not be reading and writing files constantly (e.g., generate a minute of audio, play that, and generate the next to be played)

I guess you could bore up some end-user programs like timidity or schismtracker, or libraries of the same, such as libmikmod, if you want sample mixing in midi/tracker-style.

Have you thought of looking into Pure Data? You can run it on iOS/OSX and PC - and with that you could synthesise sound on the go without having to write anything. You can, of course, blend it with wavefiles if ya like.