mr_phoenyxx opened this issue on Feb 19, 2014 · 78 posts
bagginsbill posted Sat, 22 February 2014 at 7:20 AM
Again, the blinn is most accurate. However, the anisotropic is more correct in a certain special case - a surface with repeating or semi-regular parallel groves. Anisotropic is better for nylon, brushed metal, and a few other special cases.
I wish SM would create an anisotropic blinn, and an anisotropic Reflect node. Then we'd have an easier time with brushed metal.
Phong (#2) is mathematically the simplest and the most wrong. It's reason for existence is nothing more than to let you keep using 20 year old shaders that used to matter for performance on 386 and 486 CPUs. Professionals haven't used this model since the 80s. Just forget it.
Glossy (#3) is also a cheat for special cases from the 90s. When all your lights are point sources (having no width or height) then the speculars are never looking like you get from large, close light fixtures. The glossy node fakes it - it can pretend that your infinite or spot or point lights (all of which have no width or height) are actually circles. With the advent of area lights in the 90s, this model became pointless. With the advent of real ray tracing and real-dimensioned lit objects or environment sphere (i.e. now), it is even more pointless.
Specular (#5) is OK, but really why bother. Blinn is the right one.
Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)