It's approximation of subsurface scattering, modeled by combination of multiple Gaussian blurs done in texture space (there are 4 blur layers plus original normal mapped Blinn-Phong layer).
If you are interested in details, it's basically a simplified implementation of Nvidia's advanced skin rendering presented at GDC 2007:
Is there a review of the state of the art in skin rendering somewhere? Maybe I'm wrong, but wasn't it concluded several years ago that for realistic skin we would have no option but to use multi layer subsurface scattering? (honest question, I seem to remember that the last time I read about this around 2007-2008 the consensus had just about converged to that conclusion).
It's approximation of subsurface scattering, modeled by combination of multiple Gaussian blurs done in texture space (there are 4 blur layers plus original normal mapped Blinn-Phong layer).
If you are interested in details, it's basically a simplified implementation of Nvidia's advanced skin rendering presented at GDC 2007:
http://developer.download.nvidia.com/presentations/2007/gdc/...