I've been curious about this too. Using the FOF opcode for granular sounds like the real deal, and is the common practice in Csound...but going the other way, and starting with buzz or abuzz and filtering is theoretically possible; after all, our voice is not really doing granular synthesis, but subtractive multi-filter synthesis.

My question is: what filter to use then? Resin? Butterbp? Will they just give similar results that are recognizably formant-like, but colored differently?

My admittedly brief experiments in this area didn't work as well as FOF, but maybe someone has working code to share.

AKJ