Wrong metal reflections not using a probe
under review
LottaLag Resident
Create Sphere, Open PBR Material, Make Base Color White and ORM White, Change Base Color Tint to a yellow tone, Adjust roughness to 0 and Metalness to 1. This should give a gold like material, the reflections facing the camera look correct, however towards the edge the reflection gets blueish (sky color). It appears like there is a fresnel reflection (blueish) on top of the metal reflection (yellowish). Which is physically incorrect for PBR Metal materials. The reflection on the gold material should be yellow all over, also at the edges. For reference some pictures of RL christmas baubles. For completeness' sake i added a render of a gold material in Vray (5th picture), please notice how there is no blue edge, even though the background is a 360° blue sky environment.
Log In
Dan Linden
under review
Dan Linden
Hi LottaLag.
So we can get a good comparison, what EEP setting are you using in the screenshot? If you happen to have an example rezzed in world, what is the location?
LottaLag Resident
Dan Linden Hello Dan, i am using the standard daylight eep, i am not using any cutom made eeps. The blue fresnel reflection is only visible at daylight it seems, its best visible with the A-12pm and A-3pm daylight. The location is Hundertwasser <92.400, 217.798, 2399.403>
Crexon Resident
From the pics it looks like you are trying to compare one that is outside with bright BLUE skies, to a IRL pic that was done indoors using WHITE studio lights.
A more accurate way to compare would be to use a gold material from someplace like AmbientCG https://ambientcg.com/view?id=Metal034
Switch it to shading preview and have it set to Studio environment.
Then you can grab the HDRi environment used there to import into SL to have like for like environments
LottaLag Resident
Crexon Resident i know that the images are in white studio lights, but gold will never have such a blue outline even when out in the open. Its not how metal reflections work, and if you put the same material inside a probe it looks completely fine.
The goal of PBR is to make things look more real, and not more like CG. So i am not seeing the point of using CG examples for proving my point.
Anyhow it would be great if one doesnt have to use a probe for every piece of metal that isnt the main attraction..Below is a golden ball shot in the open, as you can see there is no blue sky on a yellow metal at the edges:
Crexon Resident
LottaLag Resident its sounds like your beef is more with the glTF PBR spec which is not a LidenLab product, thats outlined by Kronos group.
LottaLag Resident
Crexon Resident i dont know how linden lab was programming this shader, to me it looks like there is a "fresnel" added to the metallic reflection. A yellow metal like gold would filter most of the blue components of the environment out.
Anastasia Horngold
I actually do see a blue edge on the image on the left. Not as severe as your SL examples, but still there.
LottaLag Resident
its different in Vray, Mental Ray and Arnold for example. And there is no physical reason why absorption color should change in plain metals based on viewing angle.
If you use a reflection probe the reflection doesn't have the fresnel part in sl, so at least one approach is wrong without a doubt.
And for reference just check an rl metal, preferably a colored one.
Frionil Fang
LottaLag Resident However it's different in those other programs I have no idea as they're commercial ray-tracing products, and I also unfortunately don't own large orbs of pure copper or gold to check to make a RL comparison. I know some renderers allow a specific edge specular color for Fresnel, but the glTF specification says the edge color tends to white for metals as the angle becomes more grazing, see https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html section B.2.1.
I can see the untinted Fresnel in SL even within manual reflection probes, e.g. a pure blue sphere within a scene with warm-toned/orange lighting. The Khronos glTF sample viewer also displays untinted Fresnel on goldish metal https://github.khronos.org/glTF-Sample-Viewer-Release/ - for example in the MetalRoughSpheres scene, with the display->active environment set to Studio Neutral. I just can't find discrepancies between the chosen standard and actual result here, whether it's ray-tracing quality or not.
LottaLag Resident
Frionil Fang I could agree to the point that metal getting whiter (unsaturated) at the edge, but surely not blue like on my example, please also see the rl examples above (christmas baubles). The Blue tint just doesnt look right to the point its annoying and if you dial up the metalness you will see the fresnel part will not change or disappear, its just staying ontop of the 'metal' reflection, which in turn tells me theres an error in the shader somewhere.
LottaLag Resident
No its just no colored metal reflection looks like this, besides in sl.
Frionil Fang
I think that's how it's supposed to work, Fresnel reflection is a different mechanism that is just total reflection, no absorption involved?
Or to put it differently, if SL has it wrong, Blender does too: the attached image is a perfectly smooth, perfectly red (255, 0, 0), perfectly metallic sphere against a perfectly blue (0, 0, 255) backdrop, which still Fresnel reflects the blueness of the "sky" at low angles of incidence.