No debate here please, just vote. I'm looking for average users for each workflow. Doesn't have to be in production, just what you use most often on an everyday basis.
I couldn't tint spec color of non metals (Cloth, skin, etc) on stylized models with metalness. Colored spec reflection isn't needed for most things realistic though (except fringe materials).
Comparison above by @EarthQuakehttp://www.marmoset.co/toolbag/learn/pbr-conversion That second point in the cons of the metalness is why I will use metalness for a realistic model and instead spec color map for a stylized model. Can I still call the latter PBR because the diffuse map is shade-less with no light info?
Metalness not by choice but because it's what UE4 supports and that's really the only engine I use personally, and for work (mobile apps) my packed metalness shader is cheaper memory-wise than spec would be.
@omar3d yep even if your values are very unrealistic it's still all """PBR""" just not matching any real world thing. PBR being in super quotes because of the bajillion shortcuts taken for realtime that result in energy loss.
Edit: A con inherent in both workflows that's worth repeating is that it's mathematically impossible to get a linear transition from a dielectric to a conductor without using layered materials. So rust to metal transitions will always lose energy if you're using a single spec or metallic material (Long time coming @radiancef0rge but I did eventually get around to making a unity program showing the difference).
i prefer reflectance and it's also the model our engine uses but I've set up our substance library to support both models simultaneously with switchable outputs because I always assume the worst will happen..
Have to work with metalness approach. Personally would prefer specular one . Imo Metalness variant of PBR is most weird "improvement" I ever saw in gamedev. It's like a disaster you should adapt and survive . Still fighting its artifacts and locked nature every day.
Those "illogical reflectance values" are usually pretty obvious . And very subtle deviations don't make perception of materials less real while allow to work around some metalness approach oversimplification.
Replies
Colored spec reflection isn't needed for most things realistic though (except fringe materials).
Comparison above by @EarthQuake http://www.marmoset.co/toolbag/learn/pbr-conversion
That second point in the cons of the metalness is why I will use metalness for a realistic model and instead spec color map for a stylized model.
Can I still call the latter PBR because the diffuse map is shade-less with no light info?
@omar3d yep even if your values are very unrealistic it's still all """PBR""" just not matching any real world thing. PBR being in super quotes because of the bajillion shortcuts taken for realtime that result in energy loss.
Edit: A con inherent in both workflows that's worth repeating is that it's mathematically impossible to get a linear transition from a dielectric to a conductor without using layered materials. So rust to metal transitions will always lose energy if you're using a single spec or metallic material (Long time coming @radiancef0rge but I did eventually get around to making a unity program showing the difference).
Those "illogical reflectance values" are usually pretty obvious . And very subtle deviations don't make perception of materials less real while allow to work around some metalness approach oversimplification.