From my experience that so called physically based approach in games while does work for simple things is making more problems than solutions.
Once you need to do a material that in reality has an anisotropic highlight spot and you can't do so in game, or for some reasons your shaders have to be optimized for more performance + very basic indirect illumination with a single hdr spherical harmonic for a whole scene, once you come to this all that physical correctness starts to make you problems and the old good non-correct tweaking returns.
Only good idea is light-conserving thing so reflective material should get darker albedo, and even more darker if you decided to use fresnel . But it has been well known long before words "physically based" appeared. For many things that "correctness" turns just into redundant real time calculations while looks no better
Imo its still impossible to make everything looking nice in a game with a resonable fps using only physically based approach and with current implementation just imposes more restriction on an artist than helps
That said, if you want to find a legitimate in-game user-case where you'll be looking at any texture that close up, you're welcome to post it.
Looking at things under a microscope doesn't represent actual end-user experience.
(but yeah that looks like a basic limitation that kinda sucks)
It gets visible with lower resolution textures as well as if you zoom in a lot. :P
An option is to compute the conductor / non conductor separately and composite them in the shader rather than using the filtered result from the texture sampler as input for the lighting model.
The artifact is the filtered reflectivity which transitions from dielectric to conductor through the misty realms of healing crystals and gems (semiconductors).
It goes away if you use point sampling in the texture but of course you have the nasty point sampling. In any case the compositing option is fiddly, needs more input textures and is more expensive. So boo hiss to that.
Replies
Looking at things under a microscope doesn't represent actual end-user experience.
(but yeah that looks like a basic limitation that kinda sucks)
Once you need to do a material that in reality has an anisotropic highlight spot and you can't do so in game, or for some reasons your shaders have to be optimized for more performance + very basic indirect illumination with a single hdr spherical harmonic for a whole scene, once you come to this all that physical correctness starts to make you problems and the old good non-correct tweaking returns.
Only good idea is light-conserving thing so reflective material should get darker albedo, and even more darker if you decided to use fresnel . But it has been well known long before words "physically based" appeared. For many things that "correctness" turns just into redundant real time calculations while looks no better
Imo its still impossible to make everything looking nice in a game with a resonable fps using only physically based approach and with current implementation just imposes more restriction on an artist than helps
It gets visible with lower resolution textures as well as if you zoom in a lot. :P
An option is to compute the conductor / non conductor separately and composite them in the shader rather than using the filtered result from the texture sampler as input for the lighting model.
The artifact is the filtered reflectivity which transitions from dielectric to conductor through the misty realms of healing crystals and gems (semiconductors).
It goes away if you use point sampling in the texture but of course you have the nasty point sampling. In any case the compositing option is fiddly, needs more input textures and is more expensive. So boo hiss to that.
Anyway, enough talking from me. Good thread!