Plausible specularity values

[anon72338821]

In the thread Albedo value of snow - How to get 80% white in Blender I had the quest to search for reliable albedo values. And this time I search for a way how we can find reliable specularity values. (Specularity as in the specular socket in the Principled Shader node).
Here a list of what I encountered in regards to specular values:

• Often they say: Most materials have a specularity of 0.5. You can leave it on 0.5.
• In some tutorials they say: don’t use specularity texture maps in the Principled shader node: that workflow is not suitable for the principled shader node. Mainly use: albedo (*AO), roughness, normal, and that’s pretty much it.
• (When you bump up the intensity of a normal map, the material get’s “rougher”. Nothing to do with specularity, but you might need to adjust the roughness values?)
• When you realize that albedo values should be much lower than you used to, you will start to use lower albedo values. But then the ratio: albedo:specularity, is changing: your materials look now much glossier, less saturated (dielectric), renders look more washed out (because of higher ratio of colorless reflections.
• Overal, my impression, when using plausible albedo values (that is much lower than usual), the amount of relfection (as in ratio albedo/specularity) is quite high. So then I tend to lower the specularity to maybe 0.25 or 0.33.

If you recognize the above, my question is: How do you deal with specularity values (or how to get reliable specularity values)?

[Patrick]

For a physically correct material the specukar value depends on the IOR.

https://docs.blender.org/manual/en/latest/render/shader_nodes/shader/principled.html

Look under “Specular“.

[anon72338821]

I see there the formula:

In an attempt to put that in nodes:

I think I didn’t do that correctly, because the materials have now very high specular reflections.

[StrayBillie]

Fresnel node outputs Fresnel weight not IOR if I understand correctly. It seems you need a simple Value node there

[anon72338821]

Ah, was indeed the trick.
Any Idea if we can plug in (good) Specularity maps into the value node?

[bandages]

@anon72338821

Any Idea if we can plug in (good) Specularity maps into the value node?

Sure you could. Ones that you made yourself, not ones that you downloaded. Typical specular maps are not made for the same meaning of “specular” as principled BSDF is using. You’d probably want to remap the IOR values in some fashion to use 8 bit per channel texture images (0,1 range.)

Typically, what we see causing specular reflection is less the material, and more the thin layer on top of the material-- for example, with skin, we see the specular reflection not off the skin, but off the layer of oil that sits on top of the skin.

Note that one of the things that causes principled BSDF to stop conserving energy is using a specular value greater than 1. So even though the linked source says that values above 1 are allowed, they’re better avoided. I don’t believe the specular field is used for materials with metallic 1.0, and there aren’t a lot of non-metallics where you’d want to use specular>1.0. The ones that are, like diamond, are probably going to be tricky to get right using just a principled anyways.

@anon72338821

Overal, my impression, when using plausible albedo values (that is much lower than usual), the amount of relfection (as in ratio albedo/specularity) is quite high. So then I tend to lower the specularity to maybe 0.25 or 0.33.

If you’re going for PBR, and using actual, real-world IOR for your specular, subjectively too much gloss is a good indication that you have an insufficiently detailed normal/displacement map. You’ve already noticed that higher bump intensity makes the object rougher (but there are more factors to this than just intensity.) Higher levels of detail will reduce the apparent glossiness of the material.

[CarlG]

@bandages

good indication that you have an insufficiently detailed normal/displacement map.

Agree on displacement but not so much on normal/bump. Normal/bump doesn’t affect shadow/masking term which would affect would much light reach the point. Instead the point would be able to “see” all of its environment, showing up brighter than it should. Think of a groove in a tire. The deep would “see” mostly the vertical dark walls (inter reflections) of the rubber groove, not every part of the environment. When you get a lot of these denser than what can be resolved, you get a “darker surface appearance”. The rubber is still 1.45 IOR/4% reflective, but using this value as a theoretical truth is senseless on simplified geometry.
People should use the controls to match what the see in the real world. Don’t be too theoretical about it. It’s the same thing with specular tint - sometimes due to lack of actual surface geometry/strands, the “internal transmissive effect” where absorption color is “picked up” can be simulated with specular tint - even if the rule is “all dielectric reflections are white”.
It’s called artistic control for a reason. Use it.

[anon72338821]

@CarlG

Don’t be too theoretical about it.

Definitely agree on that, but after having some reference points, or somehow a framework as starting point. (I want to have reference points for; Albedo values, specular reflections, and strength of light).
Indeed, I see many non metal materials having tinted reflection; Satin, hair, some plastics, and I believe many more. Got the impression that there is a small amount of tinted reflections to find on quite some organic materials.

[bandages]

It’s not the occlusion. It’s the simple fact that when you hit a surface with specular, and that surface is mostly planar, you get a big broad specular highlight that is obviously wrong. The specular highlight is primarily a function of the orientation of the normal in respect to the camera and the light.

This is something you can see just by using a principled on a normal mapped plane vs on a flat plane. Specular on the flat plane looks ridiculous.

[CarlG]

Albedo values - use a pbr cheat sheet. Specular, start with 0.5 - it corresponds to IOR 1.45’ish and is adequate for any/most dielectric. Going 0.4 or 0.6 is not what is going to be what makes or breaks the appearance of the material. Satin and hair would be two good examples of specular tint towards base, because their actual geometry is far more complex than what we simulate with simple surfaces. Plastic? Not that I know of, I would use white. Many more? Not that I know of. There may be other processes at work that tints reflections (not necessarily towards base), such as thin film coatings, iridescence, or hazy gloss, but for dielectrics the “rule” is keep it white, and know why you’re breaking it.

@bandages

Specular on the flat plane looks ridiculous.

Not sure what this means. On a flat plane (for simplicity) on creviced floorboards, I can’t just set diffuse to black for the crevice, that would still make it reflect specularly. I need to reduce specular to force it to appear dark. Just increasing roughness doesn’t do that.

[bandages]

@CarlG

Not sure what this means.

It means this:

Its color is 0 black, but it doesn’t look remotely black, because at this angle, the whole plane catches the specular. And at some other angle, it doesn’t look right at all, because the whole angle doesn’t catch any specular.

But if you bend the normals throughout the plane, it stops catching the specular all at once, and the specular no longer looks overwhelming.

If somebody hires you and says, “I want a black material,” and you show them that screenshot, they’re going to say, “I said black.” And if you say, “Well, it is black, that’s just the specular you’re seeing,” then they’re going to say, “Then there needs to be less specular.” Even though at 0.5 specular, this is a very reasonable value for most materials.

But the real reason it doesn’t look black isn’t the specular value; the reason it looks fake is because it’s bizarrely flat. Things in life aren’t flat like that, so they don’t reflect light like that. But if you use low poly objects with insufficiently detailed normal maps, they’re going to look flat. And it doesn’t matter if it’s a plane or a low poly rock or a low poly character.

Now, after you make it to that step, then you can start worrying about adaptive subdivision, and true displacement, and occlusion, and all that. (Plenty of people don’t, because they need to get a few thousand frames of animation rendered in a reasonable time frame, maybe even planning PBR textures for real-time rendering, and they’re not wrong.) But the normals absolutely do impact the subjective feel of the specular.

[pixelgrip]

Lower the roughness should help in this case.Now it seems it is at 0.5. This makes the reflection diffuse.Depends of course what material you want to create.And the appearence is strongly depending on lighting.

What could help also in this case is microroughness setup or Fresnel fall off curve.

[CarlG]

@bandages

Even though at 0.5 specular, this is a very reasonable value for most materials.

It is a very reasonable value for most dielectrics, sure. But it doesn’t account for the darkness associated with loss of reflections due to material structure. Here is an example; actual displacement on the left with spec 0.5 in a white furnace, bump mapped on the lower right with spec 0.5 which makes it far too bright due to lack of structural occlusion, and if spec is reduced to about 0.05 in the upper right makes it appear closer to the displaced version:

I didn’t include it here, but the spec lowering depends on viewing angle and the occlusion of the geometry. In this case, since it’s just a 2D bump effect, there would be no occlusion going on if viewed directly from above. Therefore, don’t rely on simplified theory, observe the real material.
For rough specular, this used to look better before, when excessive normal modifications went black. Now this problem is fixed, but that also leaves rough specular go too bright because the normals are “flattened” to eliminate excessive normals.
If client wants less specular, you reduce specular, don’t argue with your client. It’s an artistic control, use it. Anyone dissected game assets and looked at them in intended engine? Do they force 0.5/4% on all assets in engines that allow modification?