There's an interesting article in SOS this month on dynamic convolution.Originally Posted by Ritzi Lee
Apparently they've been able to adapt the convolution process to model sounds which don't behave in a linear way, which means they can model anything.