Yes, just without the darkening problem at high roughness (check bottom vs. top row of the image).
Discussion related to the LuxCore functionality, implementations and API.
Yes, but well, Lux's Rough Glass is only single-scattered (dark)
VS Cycles multi-scattered model
Blender Cycles's code: Cycles: Add multi-scattering, energy-conserving GGX as an option to the Glossy, Anisotropic and Glass BSDFs (D2002)
It is based on Multiple-Scattering Microfacet BSDFs with the Smith Model
Last edited by kintuX on Thu Oct 03, 2019 5:38 pm, edited 1 time in total.
cycles's glass is super neat.kintuX wrote: ↑Thu Oct 03, 2019 4:28 pmYes, but well, Lux's Rough Glass is only single-scattered (dark)
VS Cycles multi-scattered model
Maybe best course would be adding multiscattered ggx as option for the new planned transmission for disney.
I think in general disney should be a backbone for some ubermaterial. One to rule them all.
But roughness 1 ? Why ?kintuX wrote: ↑Thu Oct 03, 2019 4:28 pmYes, but well, Lux's Rough Glass is only single-scattered (dark)
Anyway, I have the feeling that you can achieve the same result using NULL material with inside a volume with very high scattering (and multi-scattering enabled). It may also require BiDir + long paths. But, indeed, a specific and dedicated Microfacet model is a lot faster.
I prefer stay away from uber-uber-one-to-rule-all shader, since a scene with lots of different materials simply chokes and even RTX cores can't do shit about it (and we get back to CPUs ). Also, how hard it is to make custom node groups (preset) with simple and fast building blocks to have at disposal (in library or in a shop).
But I do like specialized uber-shaders for cloth/textiles, carpaint, hair (missing), wood (missing), metals (just needs simplification)... in this regard I consider such ubershader as Principled or Surface shaders like the Disney's one (but updated) or introducing Autodesk Standard Surface well suited.
IMO, what LuxCore now needs is stable foundations and smaller memory footprint, but most importantly - hardcore use - standard stuff to become production proven!
Just a proof of concept.
It's physically real or at least far closer than single-scattered model. To make milky, etched, brushed glass, snow...
Ask architects why they love naturalistic imagery, putting glass wherever they possibly can. They love to play, paint, fill spaces with light
Next to metal and concrete it is most used modern material.
I believe even lacilaci and others can confirm this is one big flaw and huge PITA for archviz artists. Also for Corona users.
Yes, dedicated one (or Cycles like implementation) would be more than enough, at least for starters.Dade wrote: ↑Thu Oct 03, 2019 5:59 pmAnyway, I have the feeling that you can achieve the same result using NULL material with inside a volume with very high scattering (and multi-scattering enabled). It may also require BiDir + long paths. But, indeed, a specific and dedicated Microfacet model is a lot faster.
I believe this would fix the darkening issues for fluids in a glass. Like juice, water... which would perfectly complement newly added caustics and SDS paths. And other volumetric/SSS renderings would also benefit since surface doesn't get dark!
I would like to see less and less pre-processing rather than another "step" to add I also don't trust the recommended values anyways.
Sharpening would be nice, luxcore renders are often somewhat blurry and I need to always sharpen in post.