Page 2 of 2

Posted: Fri Apr 28, 2006 7:32 am
by chuck_starchaser
Thanks. Tendonitis.

My thought there was like, okay, take the case of self-occlusion in specularity. Ideally one would compute intersection of a mirror-reflected ray with the geometry, for each pixel. Forget it, of course. I assume what you're doing now is make a statistical approximation by using bakings from 6 directions, per pixel. That'd be ultra-filtered. So my thought was, use 12, 16, 32 bakings, whatever, but in the vertex shader; and simply pass the final result to the PS. Just for better balance, like giving up per-pixelness for the sake of more accurate occultation. But if it's only PS3 I guess it would be no go.