"Principled v2" feedback/discussion thread

[brecht]

There are distinct diffuse and specular layers in the Principled BSDF. The roughness of the specular layer is controlled by Roughness, the presence of the specular layer by IOR.

Every shading model is an approximation in some way, so it’s hard to talk about correctness in an absolute sense. It’s just a good fit for many real world materials to have these two layers.

In this case, the diffuse layer can be better thought of as subsurface scattering with a very small radius, rather than a specular layer that is very rough. It’s modelling the volume just below the surface more than the surface itself.

[Gurra]

Hey, wasn’t the Radius in Principle BSDF supposed to be changed from XYZ vectors to RGB or did I remember wrong?

[FreeMind]

And shouldn’t have the “Radius” and “Scale” labels switch places?..

[nacioss]

I’m trying to render grass.

Been collecting some info from the internet about PBR values for Grass blades/leaves.
Found these:

• Tall Wild Grass 0.16 - 0.18
• Short Green Grass: 0.2 - 0.25
• RGB [0.080219, 0.09084, 0.05127, 1.0]
• RGB [0.09084, 0.099897, 0.026242, 1.0]
• Roughness: 0.07
• Glossiness: 0.5
• Specular: 0.043734
• Translucency: 0.1 - 0.3 (yellowish color)

Writing this because was wondering whether these are “good” PBR values, and also to help anyone looking for these resources.

[jonlampel]

This looks nice! I think it’s a good thing to mention here because really translucency should be mixed with the diffuse under the glossy layer like SSS is, but that’s the one effect that’s not possible with the Principled BSDF. I’m assuming that’s something that will be solved with the future Thin Sheet mode?

[Alaska]

@jonlampel

but that’s the one effect that’s not possible with the Principled BSDF

Users always have the option to recreate the Principled BSDF from smaller nodes, and make adjustments as needed (E.G. Doing the diffuse/translucent mix under the glossy layer inside the custom Principled BSDF).

However I should note that this is only possible at the moment using OSL due to some of the features required to recreate the Principled BSDF from scratch missing from standard nodes. Work is being done in this area to allow users to create the Principled BSDF from smaller nodes without having to resort to OSL, but the timeline on when that will be finished is unclear/or if all the changes necessary will be accepted, is unclear.

[Eneen]

I think thin model is on the TODO list

[mr.A]

Hi guys,

I’m new here, so i hope this is the right thread:)

Just one question, as this behavior differs from 3.6 to 4.0

There is one area light from the left side of the sphere.

Sphere BSDF settings:
Base color- black
Metallic - 1.0
Specular tint - black

Is that correct that the reflection from the area light is still visible (4.0)? Its strange, as its visible only on the edge and disappears if i rotate the area light around the sphere closer to camera axis. In 3.6 there will be no such edge reflection in this situation.

Thanks in advance!

[Crayz_Crinkle]

I really REALLY hate that they removed transmission roughness. That was vita to making glass or water with clear reflections, but blurriness of objects within or behind the see through object.

Please please for the love of god bring this back!!! Without it, you can’t make transparent glass or water materials.

[JulianPerez]

If you use the Principled BSDF you can use the Coat to add a clear reflection to the material on top of the main roughness

[Bobo_The_Imp]

A lot of people seem to be saying that using roughness then compensating with coat over the top is a good replacement for transmission roughness, but I do not agree.

I have made a chart here with Renderman, Cycles from Blender 3.6, and Cycles from Blender 4.1. The material I’m trying to make is a foggy glass, rough in the center, shiny on the surface. In Renderman I am using transmission roughness, in 3.6 I am using transmission roughness as well, and in 4.1 I am using roughness, coat with a weight of 1, and coat roughness of 0. All blender renders are with full global illumination settings (32 samples on everything)

Renderman seems to conserve a lot of energy, while 3.6 seems to lose a lot within the glass, one might be more accurate than the other, I am not sure, but they are a good comparison despite renderman being much brighter because the look of what the transmission roughness is doing almost exactly the same thing between the two. Note the darker parts of the refraction, the milkier, foggier left eye, they match super well!

The 4.1 one renders with coat look sort of acceptable below a value of 1 for roughness, but in my opinion there is a pretty large discrepancy.

This is a comparison between the above 3.6 Transmission roughness of 1 and 4.1 coat method. I would have marked the pictures with my observations but didn’t want to create any visual noise. The contrast between dark spots and bright spots is drastically reduced, making the 4.1 render a lot more flat and washed out. The general curvature around the area between the two eyes is actually dark in the 4.1 version, whereas in 3.6 it is much brighter, so those dark spots vs white spots have swapped in that area of the face. If we look at Suzzanne’s right eye area (remember it’s mirror lol) we see a lot of red bleeding through from the background, this is not present in renderman or 3.6.

And my least favorite:

Since the coat is on top of the roughness layer in the principled 2.0 (which is a good thing) we get a very sharp coat reflection of the window in the HDRI I used, but because the roughness underneath is set to be 1, we get a stretched out reflection underneath it. This makes it look like the coats reflection is washed out, even though technically it is quite sharp.

For these sort of reasons I really don’t like using coat as a replacement for transmission roughness. In my opinion it is a feature we just cannot reproduce currently.

I do hope in the future we get a return of transmission roughness!

[EntityDesigner]

what type of material are you triyng to do?

[Bobo_The_Imp]

Just a foggy glass. The inside is rough, the outer surface is very shiny

[Jvry]

4.1 from a distance looks like frosted glass to me, though still not quite accurate as it should have a rougher surface, which admittedly wasn’t your goal with the material you were aiming for, so could be made closer with coat roughness. The lack of coat roughness makes the higher roughness ones look ever so mildly pearlescent up close, so that’s another glass effect that wasn’t your aim.

Can’t play about right now, but I wonder if some volume options might help? Though I doubt that would be acceptable to many people looking for replacements to old workflows.

[Bobo_The_Imp]

The name of the glass can be confusing, since I’m not sure if frosted glass’ surface is supposed to be rough or glossy. Names aside, the goal is to either hit something like 3.6 with a transmission roughness of 1, or render man with a transmission roughness of 1. So top left, or middle left is the overall goal.

Ideally the way I want people to view my post above is to compare 4.1 with everything else, and hopefully realize it’s not able to produce the same effects, at all in my opinion.

If transmission roughness were added in volume options that’s just as well. I think Lucas was potentially going to add volume controls in the principled BSDF anyways.

Doesn’t matter where it goes, but would be very nice to be able to control this again

[EntityDesigner]

use the lighpath → Transmission Depth node, I think what you want to make i’ts a glass with some type of impurity in it’s interior.

[Bobo_The_Imp]

@EntityDesigner

some type of impurity in it’s interior

I suppose so. I’m not sure if there is a name for this, but the first image you have there is the general purpose of the deprecated transmission roughness.

This is a much better way of achieving this look. I would not recommend anyone uses the clearcoat method unless its very low roughness and clearcoat roughness.

Here is a side by side with the transmission depth from the lightpath node in 4.1:

It still lacks some realism, this way is very “even”, where 3.6 and renderman will probably do some calculations based on volume or curvature etc to make those nice variations in the refraction, but I think this is a better work around for now. For now being a keyword, I still think transmission roughness should be re-instated at some point, it is a very common factor in many BSDF’s amongst a lot of software.

[Alaska]

@Bobo_The_Imp

If transmission roughness were added in volume options that’s just as well. I think Lucas was potentially going to add volume controls in the principled BSDF anyways.

The main idea for volume controls is to allow you to control the colour of transmissive materials through volumetric effects. This will give you the ability to have “light fall off throughout a medium”, similar to this, but in a more user friendly way:

Weizhen worked on a prototype for this using Beers law approximation for volumetric adsorption. But I believe it hasn’t gone further because the Cycles developers would prefer to use true volumetrics: #111806 - Cycles: add distance-dependent transmission color in Principled BSDF - blender - Blender Projects

[Alaska]

@Bobo_The_Imp

A lot of people seem to be saying that using roughness then compensating with coat over the top is a good replacement for transmission roughness, but I do not agree.

They do operate differently, hence why you get different results.

If you would like to replicate the effect of “transmission roughness” in Blender 4.0+, you can use a node setup like this:

I should note that it’s not perfect. I’m missing some things from this node setup that make it a perfect match for 3.6. Mostly because some tools needed to recreate it are missing (E.G. The Principled Fresnel), or in one case, I didn’t bother to properly investigate how it worked to see if I could recreate it with nodes (E.G. How cspec0 influences fresnel).

[Bobo_The_Imp]

Ah yes, that’s better yet. Thank you so much for that!