Epic Games / Disney White-Washing

Did he? Or is it just the way you’ve interpreted his somewhat cryptic formulation and associated it with an issue you have with UE4?

He has posted 3 quite different pictures containing a very different sets of materials each set up in a very different manner. Comparing all 3, and attributing clearly absolutely different combinations of shading scenario to a single, small issue which is usually barely visible to human eye is quite a stretch.

What you are likely referring to has been known as rough/glossy fresnel: https://www.chaosgroup.com/blog/unde…glossy-fresnel , a formula made to correctly compensate for diffusion of reflective energy on the grazing angles as the roughness increases…

BUT!

I have been using offline renderers lacking proper glossy fresnel implementation for years, and once both of them finally got it (Corona and V-Ray) the difference was obvious, although not big. Non the less, I can spot lack of glossy fresnel implementation when I see one, and it certainly does not concern unreal engine. You can tell absence of it if you make highly specular, highly rough completely black material and then put it into bright environment. If the glossy fresnel compensation would not be present, then you’d end up seeing very obvious bright ring around the grazing angles, yet when you try that in UE4, you get the right result:

If Unreal did not implement any handling of glossy fresnel, you would literally see something looking like this, cause that’s exactly what I saw when I was using renderers without glossy fresnel:

So if you feel like your speculars on rough materials are too strong, then it’s most likely just your subjective unrealistic expectation of how the material should look under given lighting conditions, rather than issue with the material implementation. And if you are doing some fake things to fix that, you are most likely breaking it instead :wink:

3 Likes

Eh, all I know is that dirt and grass don’t turn solid white when looking in the direction of the sun in real life…

1 Like

In lesser times, I would say ‘QFT’, for now, you get a like :slight_smile:

I do feel (completely subjective, as Rawalanche hinted) that The Disney Problem in rendering for Unreal Engine (takes a deep breath) is not the best way forward. The glinting sheen on surfaces to give the rendering an ‘out-of-the-box’ default look’n’feel of looking like a Disney render of real world look up tables of the physical properties, is an issue.

Lest me fall-back, slightly, and say this (UE4) is not the only game engine to exhibit these qualities, as the CryEngine/Lumberyard has these issues. As does Frostbite. As does Unity. As does Stingray. As does Creation Engine. As does RAGE (the grass in RDR2 is white). As does Carbon. And so on.

Another thing that springs to mind on this topic is one hinted at by Crow87, is the LUT and following that, HDR values and how one should ultimately ‘master’ their color values from SDR to HDR.

A cheeky comment from me, would be that this is the new ‘bloom’ we got with DX9 in the Source engine and of course, UDK. In that the default view has this artefact, and game engines using PBR/Disney for their games are looking…very odd - exhibiting this ‘trait’ of glinting angle whitewashing of the color values.

I don’t think color grading and LUTs have anything to do with these issues since they can’t fix rendering problems. Sure, ACES desaturates highlights, but you have to have a pretty high scene-referred color for it to be super noticeable. Crow87 was likely referring to the Log formats most cameras can shoot in to replace Raw, which is not a format meant for viewing.

In my opinion, the first Disney shot looks fine. Maybe a little over-exposed, but the only white sheen I see is on skin and fabric to suggest asperity scattering.

2 Likes

I wonder if “normal map mipping error” contributes to the white reflective dirt issue? I don’t know if UE4 addresses this in any way.

http://media.steampowered.com/apps/valve/2015/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf (page 28)

He was not the first and likely to be not the last from whom I’ve heard about it. Different artists word it differently. Over bright-specular at an angle, wax-like appearance, white sheen, plastic look.

Maybe. Or maybe not. In view of a sentence above, I’d lean towards the latter. I do not have any issues with UE4. In fact, the matter is not specific to UE4.

Nope, I am referencing to visibility term and associated issues with using it on pre-integrated data. But the nature of complication, in response to which it was introduced in vray and then in the other renderes, is the same.

I would not doubt your experience for a split second, but glossy fresnel, as you know it, is not present in UE4. And it should not be there in the form familiar to you know from offline renderers.

If you ramp up and isolate spec IBL contribution high enough, you will see it on a rough sphere.

https://i.gyazo.com/5e84fb0e252cf1ed341d24b888c968a5.png

I claim only that indirect specular contribution for rough surfaces deviates from expectations and that it plays a role in artists complaining about it occasionally. The issue is not with specific implementation or engine, but theory behind it and I am not basing that on personal experience or, in fact, anything subjective.

The way I am doing it certainly breaks things.Dampening fresnel for indirect spec is a patchup without any theory behind it in response to artistic issues.

It does.

Well of course! If you push specular IBL contribution to the extreme and isolate just that contribution, of course you should get falloff on the grazing angles. That’s the fresnel effect!

The point of the glossy fresnel compensation (which Unreal certainly implements, but maybe in some simplified way, otherwise the results would be very different) is just to reduce amount of fresnel effect based on roughness, not remove it all together. You are complaining about seeing fresnel. It almost sounds like you are proposing to remove fresnel altogether and have just constant reflection regardless of angle.

If you crank up IBL specular contribution and isolate it to see the fresnel effects, then that’s good that you see it. It means it exists. If you saw just a flat sphere, that would be a lot more concerning.

1 Like

Yeah, I wasn’t talking about the spec calculation or the glancing angle highlights - I’m only referring to the desaturated, low contrast look described as ‘white washing’ by the OP. The two aren’t related, and the latter is just an artistic choice rather than a fundamental problem with the DPBR calculation.