PBR; is it really as much as the holy grail of shading as people say it is?

Physically Based Rendering, it has taken the CG world by storm and both game engine and render engine developers are implementing it as a full-on industry standard for materials.

However, I look at some of the examples and I can tell pretty much right away that they look CG. Noting some of the examples taken from 3D Coat, somehow it seems like they would be a little too ‘sterile’ if you did a direct side-by-side comparison with a photo. Now I know there’s the whole idea of ‘energy conservation’, but with the way it’s being realized now, images that use them seem to lose a little too much energy around the edges and glossy materials have an apparent tendency to look like they’ve been dipped in a sort of glossy veneer rather than looking truly integrated into the material.

In Cycles, the fact that the add shader node exists allows you have to have full control over the energy, and I am happy that it has stuck around, because it seems to me that the purist approach to energy conservation appears to just miss something when I look at CG and then look at photos (in Cycles for instance, just making use of the mix shader node for SSS-heavy materials can lead to ugly falloff colors in shadowed areas). Is it just me on this then or is it true that the modern PBR model doesn’t currently have all of the answers to CGI materials?

Case in point, I don’t know if I’ve ever seen a lot of cases in CG animal design that can perfectly capture the same level of realism as seen in this image

You look at it and there almost looks to be a sort of glow to the inner mouth area, I get the impression that the PBR paradigm is trying hard to prevent that from happening, so you can more contrasting within the shadowed areas than there should be. I’ve done a number of Google image searches for CG animal renders and it’s not easy to find one with materials that give that final spark it needs to look truly alive.

I guess PBR is a simulation-like paradigm which in theory should yield more realistic results than other “fake” methods.

However, a simulation is only as realistic as the assumptions that go into it. I get what you mean with things looking pretty real, but not really-real.

A realistic render depends on all details being realistic. I think that, as always, PBR realism depends just as much on the artist taking the time to make ALL details realistic, even if they use (faster?) PBR methods to create those details.

From what I have seen, Cycles or other PBR systems let the minimum effort produce an OK result, but there is no shortcut to a good result. Or, since people have started learning cycles, the average crappy render has gotten better, so has the average good render, but there are still only a few GREAT renders.

Also, good nature photography often uses long lenses, or even flash… and you need to have proper SSS settings on the mouth area to get a look like that… my guess is not that PBR is the wrong approach for this, but that it’s very rare for any artist to go to the trouble to recreate the right lighting conditions and internal structure of the flesh of the mouth to create that “really real” effect…

tl:dr “it’s not the tools it’s the artist”

PBR is only really a new thing in the realtime gaming world. Because we finally have enough computing power on the consoles to support it. Yeah, it’s mostly about energy conservation. But it’s also about using reflections rather than just white blobby specular highlights, and about avoiding nonsensical surfaces, like people putting colorful specular maps on dielectrics.

The trend is mostly about hard surfaces. Hair and skin, like on your hyena image, don’t really apply at this point in time, at least not in realtime - it’s hard to talk about energy conservation when your SSS shader is just a screenspace fake. Same thing with hair.

I can tell that PBR is likely a step in the right direction, and images such as those in Allegorithmic’s PBR gallery shows that.

In that gallery though, it seems like most of the creature models don’t have quite the right balance between SSS and glossiness. Some are really shiny while others look a tad too diffuse or a tad milky. It’s a great advance for realtime graphics, but the impression I have gotten is that some want the exact same model to be used for general rendering of images (or even more so, have it actually replace the existing shading system in some 3D software).

As for images, has anyone managed to use the Disney Principled BSDF for photorealistic images yet, it’s considered to be the base of some of the existing systems and we have already seen in film that it is awesome for toons, but what about matching real-life?

PBR is really interresting I think !

The ability to tweak a node in the viewport is really great and speed up the workflow.
That remember when I worked on Guerilla render, we had the ability to solo a node and view it in the viewport, so, no need to start a render, It was really fast and cool.

If PBR is well integrated in blender, we could paint/tweak our shaders without the need to start a render and we will have the same result on the render.

I made this drone in blender, paint it in Substance Painter and render it in Cycles with the PBR shader from Mattrm.

For characters, it’s up to the artist to make great shaders.
The users of painter/designer are mossely game artists, we just need to wait for better renders.

Blender needs PBR and I hope we will have it as soon as possible.

The concern I have though, is that some want to completely replace Blender’s shading system with this (including the different shading components in Cycles).

The nice thing about Cycles is that, as you add more shading components to complex cases, there is an easy way to compensate for possible energy loss. I recall a thread from the Octane forum a while where someone was struggling to make more energy available to his shader’s SSS component because the other parts were taking a rather large share. If the energy budget is making it hard to get what you need among the different components, you just expand it.

Now to mention, I actually wouldn’t be opposed to a separate PBR node for Cycles as a sort of ubershader, as it will leave the existing system alone if you rather use that (or mix them together).

Yes. Yes it is.

Can you please tone down the clickbait style headings Ace Dragon?

Is PBR the holy grail? For this generation and some future generations…absolutely. The future for real time graphics, well graphics in general actually… is not in the mesh work or how many polys you can push… but in the shaders. Shaders are the “future” (for the foreseeable future, always open to change). This is the basis in which Allegorithmic has built their substances off of… they an be dynamic and include all sorts of information. It really opens up the door to many more possibilities, and efficiencies as well in terms of cost (cpu/gpu).

That said, because its still new what we see is just the making of a foundation, a standard. A lot of artist dont even know how best to maximize and leverage the new approach, so it takes some time for this tech to adapt and the artist to learn all the ins and outs of this method.

Look at this Ace Dragon !

No need to change Cycles nodes :wink:

I thought cycles was PBR… scratches head

Anyways, PBR to me is just like a brand new ice cream flavor. you get excited about it the first few days, but eventually you’ll return to your real favorite.

yeah, yeah, a lot of new video games use PBR… but it’s really more of a marketing tactic than anything else. it doesn’t add much to the experience. not to mention, a lot of the great games out there are simply not PBR-material

Quoted for agreement - this is highly liberating in the texture workflow for feedback, and +1 for the link Pitiwazou shared there for the nodes

In this case, I think it’s more appropriate as people around the web have been hailing this as one of the greatest things to ever happen to shading technology.

I thought cycles was PBR… scratches head

The materials are physically-based, but Cycles wasn’t specifically built around the modern PBR paradigm, it’s a lot like Luxrender and a number of other pathtracers that had their material systems developed before it really started to become a thing.

Ace, I’m really not sure you understand what you’re talking about here. Do you know what the node setup looks like to reproduce the basic UE4/Substance Painter shader in Cycles? It’s this (you can drop the second mix/glossy if you don’t need metalness):


How about LuxRender?


These “PBR” shaders you see in tools like Substance, Marmoset, UE4, etc, are just uber-shaders made out of things you’d find in Cycles, Lux, Arnold, Vray, or any other modern offline renderer. The reason this is mostly a big deal in the gaming world and not on the offline world is because none of this is new when it comes to offline raytracers. The original PBRT book is over 10 years old now. People have been making films and commercials with “PBR” for years. We only got this “PBR shader” hype because the realtime folks decided to semi-standardize some ubershader conventions like the albedo/rough/metal thing when implementing these techniques in their tools.

It’s mostly about things like energy conservation and IOR (view-dependant reflectivity), which were previously not provided by default shaders. You could set them up yourself ofcourse, that’s true, but they weren’t common in gaming.

Together with PBR, there was a push for other things that, --while not directly called for by PBR-- are very beneficial to consistent, ‘realistic’ lighting.
For instance: Linear shading, image-based lighting, blurry reflections and a different set of imputs, including Metalness (depending on workflow) and Ambient Occlusion (no longer directly multiplied on the diffuse or spec). Sometimes even screen-space local reflections, which in my opinion greatly aid the look of PBR.

Here’s a good reference for those who are interested (the Physically Based Rendering Bible):

edit: also, Ace? The clickbait title wasn’t called for ‘in this particular instance’, anymore than it ever is. Your posts rarely strike me as helpful or informative, and you seem especially adept at baiting people into arguments.

Okay So lots of confusion here about pbr…

pbr is pretty much only relevant to real time (gaming) industry and blenders viewport(for making gameart, think glsl viewport replacement). You will not be using pbr to preview cycles materials you should still be using the already provided realtime render preview.
PBR still does a lot of faking compared to render engines like cycles or even Blender Internal. for instance: with reflections, in pbr it’s all baked in via sphere map capture onto the models lightmap, global illumination is still too costly to do in realtime so you still have to bake out an ao map for each model and ao doesn’t give an absolutely correct global illumination look.

No, you can use pbr shaders to paint your model and use it in cycles with the same result. I did it and it works pretty well.
PBR is not only for real time, it’s also for precalculated renders.

It may be worth noting that all PBR is an approximation anyway… at the atomic/photon level, everything is either reflected or absorbed, and materials take on different looks depending on the microscopic shape of the surface and the configurations of the electron orbitals of the atoms that make up the surface, or some shit… now THAT is physically-based…“roughness” and non-spectral RGB colors are a very cavalier interpretation of actual interactions of photons with physical objects.

They just happen to be useful approximations that work on a macroscopic scale for a decent range of materials.

PBR is not some magical equation that makes everything photorealistic. It is however a big step on the way for realtime graphics but until we have hardware that can accurately render multi-million bouncing lights with wavelength and true physical materials in realtime we will not have true photorealism in games.

Do remember that everything we do in realtime graphics (including PBR) is just an approximation of reality and as such it is just an elaborate fake.

Photorealism requires no such elaborate simulations. Pictures are photorealistic when the intended audience can’t tell them apart from photographs more than 50% of the time.