Hey, I somewhere read about this hack that for SSS you first render a depth map with all back-facing polys, then another depth map with all front-facing polys. merging these maps gives you an idea of the geometric "thickness" of the underlying geometry. how can I further exploit this information to simulate SSS?? thanks, Nick

Read up here on the full thing: http://download.nvidia.com/developer/presentations/2005/SIGGRAPH/Truth_About_NVIDIA_Demos.pdf

In few words, it's a measure of the amount of light that goes through, but it is only a coarse approximation, or a hack if you want it that way. A more or less ok looking factor that you can multiply your backlight term and "dense structures" map with.

You need to perform a blur on the "thickness" to account for the shortcomings. First, what you get is jaggy and not terribly accurate (unless maybe if you use R32), and second, it is not a real measure of thickness, but the distance from the closest to the farthest point (whatever they are). This may accidentially be the object's thickness, but can as well be something completely different (for example, imagine your teapot's handle was pointing directly into the camera). The blurring helps to "iron out" the crinkles from the method.