Hi there!
Lots of fine and talented ladies and gentlemen here, i feel i need to finally stop lurking and get my concept together.

I'm banging my head around a concept of realtime granular modulation to the incoming signal.
Every granular synth i have encountered is built upon samples - either recorded or read from a disk.
I would like to make grains out of heavily timestretched first couple of ms of the incoming signal and then continue with the rest of the signal, probably fed from a continuously refilled buffer.

As I'm only a humble piano player the idea of making two tables that somehow manage to synchronize the out-spitting of the little grain-building chunks of audio makes me wonder if there isn't any widespread solution to my problem?

my signal would need to be timestretched at least 8 times its size, so I think there should be plenty of computational time to spare.
For added value I plan to implement beat synchronization so on every next beat the grain catches up on the incoming signal (scraps or saves the buffer and fills it anew)

So putting a long story short: how to feed a realtime signal into a variable length grain window, saving the residual realtime signal somewhere else?
should the timestretching be constant or is it possible to modulate this also?

I hope i made myself clear and someone will offer a helping hand so I can start coding. Of course if I succeed I will share the results

Nikola Kolodziejczyk
http://music.thinkbay.net