Megascan does around 0.83. Some guides tells it's 0.76. Chat GPT told me 0.75. Adobe substance libraries have a lot of around 0.5-0.6. Textures.com are in range of 0.75-0.9. It all gave pretty subtle difference in our old renderer and Megacan ones looked almost matte and now 0.76 looks like a nuke when the sun is around 25 degree above the skyline .
In offline renderers it's not a big difference especially with ACES or AgX tone mapping working . In our new renderer I see very strong intensity of highlight with low sun and hardly anything when the sun goes 50 degree above the horizon with 0.83 while with Cycles, Arnold or Octane it's still a lot of pretty noticeable wide spread highlight spot. So I wonder what you guys do?
Replies
measuring roughess and translating it into something that matches your specular model is probably all but impossible
as such - as a surface gets further away from the viewer, progressively larger changes in surface direction start to behave like microsurface information - which is what your roughness texture defines.
in short - eyeball it and have some sort of toksvig roughen over distance thing going on in your materials
other things contribute to this as well - particularly with glancing light angles
dielectric reflectance and different diffuse BRDFs to name a couple