Understanding PBR & the Principled Shader | Blender 2.8
Vložit
- čas přidán 25. 07. 2024
- Join the Squad and Share your Work on Discord: / discord
I created 5 Free Automotive HDRs for Download: mouval.com/free-hdrs/
Patreon / mouval
Website mouval.com/
business inquiries or donations :)
db@mouval.com - Jak na to + styl
You are the best person.....I am literally binge watching every tutorial of yours..............awesome.
I forgot to say thankyou so much:)
i just saw 0 dislikes.........u deserve it man
Thanks a lot! Good to read cause right before this comment I read how I am discouraging upcoming young 3D artists haha. So I do got haters from time to time. Glad you like my stuff! :D
@@DamianMathew I'm glad sir that I too couzld help you👍
@@manasagrawal2149 :D
10:10 I used to think that too but after doing a lot of searching I found out most of them think it's for specular workflow or it's not for realism but for realistic rendering you should use a specular map! Most realworld materials have specularity as 0.5 but some surfaces have less than 0.5 (wet surface has specular value ~0.2 ish). This specular value is not the specular workflow that old game engines used but an IOR value of the diffuse material. Every object has a IOR (specular) that determines the reflectivity of the surface (on how much angle you can see the reflection). BTW keep up the good work I enjoy your videos.
A3 K7 that’s quite interesting what your saying. Because I did notice specular on 0.5 being visually too high sometimes. I often turn it down a bit when it just doesn’t look quite right, but also sometimes because my HDR isn’t calibrated well. I always thought it’s a specular workflow or just an override for stylized reflections. But what your saying sounds right, I will definitely look deeper into that topic. And I’m glad you like my videos, maybe I will make a short video on this topic once I read a bit about it!
Maps on specular should be special purpose, not something you *should* do. But *do* use its slider artistically - too many are obsessed with the IOR value for some reason thinking looking up IOR will make it look better - far from the truth.
As a classical example, rubber probably has a typical dielectric IOR or close to it - 0.5 is good enough. But that makes the reflections too bright - why? Because of microscopic structure (too small for normal maps). Take a look at microscopic pictures of rubber and picture a ray bouncing around in there that will get absorbed before some of it eventually gets out. Specular value is an artistic control, use it artistically and drop the "realistic IOR" nonsense. Set it to what you observe or what your experience tells you.
Reason IOR and thus specular reduces on wet surfaces is because IOR now is computed between surface and water rather than surface and air/vacuum.
@@gottagowork True as most of the time the difference is almost not noticeable but IOR is useful for refraction e.g. when making accurate caustics/glass or even when creating physically accurate "perfect" base metals.
e.g. Gold = 1.35002 ( fbe6ab ), Silver = 1.082 ( fefefd ) w/ Metalness workflow ofc. Before Metalness was integrated I've used to create fresnel ( or color ramp in other engines ) bitmap with values based on wavelength on the specular/reflection slot of the material which was not super accurate because of the limited size of the curve editor and time consuming.
Those who are interested for more info google: "Understanding metalness" posted in Chaos website.
@@ExacoMvm Well I disagree with Chaos on this one 😀Looks kinda dated. They're using "real fresnel" values for metals when for accurate results you should be using "complex fresnel" (n/k values) - facing and edge tints instead of fade to white reflections. And real gold is actually out of gamut, so fb as red doesn't make sense - but who cares. Still, it's not something I bother with anymore until we get complex fresnel as n/k inputs or "Artist Friendly Metallic Fresnel" available in most other renderers. Nice to have, but not something that makes or breaks an image imo. Far more important to me is not relying on numbers but what I observe and "measure" and use the artistic specular value as it was meant to; most materials are somewhat porous allowing specular energy to get "absorbed" into the void (micro-holes in the structure) rather than being reflected back out at full strength as defined by fresnel. There are lots of things lacking in Cycles, but now everything is about geonodes.
Hey man! I loved this video and also the one with car-paints. I loved your content and especially the car paint thing. It was excatly what I was looking for for a long time on CZcams, but you had your comments turned off there so I came here to thank you for that video. You've earned my subscription mate.
Cheers!
Shaurya Pant oh comments were off? That strange I’ll check that! Yeah this video was way longer than I expected but I think it was necessary to understand. Cause once understood it’s really easy to almost recreate any material. Awesome you liked it and even watched it to the end haha
@@DamianMathew Yeah this was a little basic, but I guess you have checked the 'kids mode' option by mistake for the paint video... Hence comments etc. are turned off...
Shaurya Pant aaaaah yeah maybe
5 reasons to adjust or map specular value:
1. Using topcoat, builtin principled or better custom one, will reduce the IOR for the substrate. Lower plastic bubble wrap or ice cube in water and see reflection reduce. The IOR value is a ratio between material and vacuum (or air, close enough). Once the vacuum/air is replaced with another material, the IOR changes.
2. Simulating cracks or shadow gaps in a flat plane. I use it on fake floorboards all the time to simulate their shadow gaps. I also use it on cracked wood. Simply increasing roughness is not enough.
3. Very bumpy surfaces using bump or normal maps on the macroscopic level. Check what the value should be using true microdisplacement. Due to shadow and masking effects in the real displaced version, the reflection is reduced. Energy is lost for each "in-object" bounce, this does not happen when modifying normals - then it will always find the environment. Try it out in a white furnace test. Might want to modulate this based on viewing angle.
4. Macro-cavity. For similar reasons as above, but you control it with a cavity map or AO. In a game engine you would do it stronger as specular "rays" will always reach the environment rather than get dimmed down from interreflections. Game engines handle this automatically (to some extent) in how ambient occlusion map input is used.
sponge
gottagowork wow great explanation! Makes perfect sense! Maybe I’ll make an update video about this topic. Thanks for the great constructive add on to the video!
Base Color, aka Albedo - tends to be in the high range for metals and low range for dielectrics. Use PBR cheat sheets to scale or verify the texture you're using is within plausible range. I.e. a white dielectric wall is not [0.8, 0.8, 0.8], but way lower. Those kinds of albedo values belong to snow and thereabouts. Using plausible albedo values are one of the key points of PBR.
Specular Tint - can be used if faking colored hairs on a surface. These "hairs" will pick up the main color to some extent via transmission before it is sent back to you. But yeah, use it artistically.
Sheen - technically it's a compensation for energy loss due to disneydiffuse having exit IOR. Without going into details, it basically makes the rim darker, and energy is lost in a furnace test. Sheen "kind of" brings it back. But sheen is also a forward scattering component, the only one we have in Cycles - I prefer the Velvet look myself but this doesn't do forward scatter. Adding sheen breaks energy conservation more (Principled is "broken" anyway), and sheen is buggy if used with global AO.
Anisotropy - Most machined metals should use this unless polished to a mirror surface, anything made from brushed sheet metal should use this. It is something you should consider for any surface with a directional pattern in it; wood tends to be somewhat anisotropic, but we usually don't care. Even skin becomes anisotropic if deformed and stretched. Don't forget to use UV based tangents if radial tangents don't work well, i.e. on a Utah Teapot. For highly anisotropic but smooth surfaces where you can't feel the grooves with your fingernails, instead of bump mapping, you should modulate the anisotropic rotation a tiny bit. Anisotropic rotation is just the angular offset from the tangent. I never use the Principled shader if I want anisotropic effects, a manual PBR setup isn't that hard to do, and the anisotropic shader also have the Beckmann distribution available which I usually like to add in for better highlight spread control. Often anisotropic reflections tend towards isotropic reflections towards glancing angles. While it adds shader call costs, it's better to mix in a separate glossy than try to manipulate anisotropic value and roughness. On a separate note, also add separate glossy if you need to do hazy reflections.
Clearcoat - it kinda breaks energy conservation as it is simply *added* on top. This secondary specular lobe uses GTR 1 rather than GGX (aka GTR 2), and looks much better for a clearcoat with its broader tail (I want this for metals damn it!). Unfortunately it's roughness control is kinda broken, make sure it never goes below 0,02 (use a maximum math node if used with a texture map). This is a very simple clearcoat which you can use on finished wood, but as it doesn't do absorption or internal reflection effects, you cannot use it for thick coats like car paint, epoxy or liquid glass type of coats - at least not without some fakery to the albedo. Clearcoats in the real world is used to add durability, improve surface finish, and smooth out imperfections. So in addition to any bumps that are unique to the topcoat (like wear and tear from traffic on finished wood), you can add in some of the bumps in the substrate. Remember to lower the specular value to 0.1-0.2 in the main material.
Absorption/transmissive effects - do yourself a favor and learn how to setup manual PBR. The only component to this is the reduction of fresnel based on roughness to mix between whatever absorption components are simply mixed together, and the glossy component at the end. Although controlling shadows can be tricky, keep in mind Principled doesn't do this at all. You have to rely on caustics, and SSS doesn't light up other parts at all - you need to combine with translucency which Principled doesn't offer.
Translucency - it's just a diffuse shader for the backside. You use it if you want a uniformly lit backside with shadows. Always mix it with diffuse, but it's color should be darker and more saturated depending on how much color it picks up through the material. As it doesn't have highlights, you'll need to mix in rough refraction using geometry/incoming as a normal (for thin surfaces).
Great add-on to the Video, thanks a lot! :D
Jesus christ, just when you think you understand blender you read this and realize you know nothing yet😂😂
I guess I gotta keep learning
@@imyourmaster77 On the other hand, if it looks good enough, then it's good enough. Often it's only in the eye of the creator these differences will make a difference. A client might see there is a difference, but completely unable to pick out which one is less accurate. Still, it's useful to know some theoretical concepts.
The knowledge that you have dropped in the comments here is impressive. I have been doing a lot of reading about PBR, even some popular papers, but this goes much deeper than most sources I can find - at least in regard to application. How did you get there? If you can recommend a one or two good reads, maybe about the described anisotropy best practice, I'd be very thankful.
@@lemming77gode Documentation from other renderers, marmoset toolbag, the various Disney papers. Mostly those kinds of things. But from memory, so always look it up in case I got something wrong - happens. :)
As for aniso, don't know of any "best practices", except aniso in Blender is somewhat broken. Tangents are not smoothed across the seam for UV based tangents (normals are smoothed), and reporting it got the usual "it's by design" response or some crazy workaround nobody could be bothered with. Another could be related to GGX tail vs GTX tail, not sure, but we can't achieve aniso "light stretching" without excessive roughness. I can get closer mixing in Bechman aniso.
I heard, that Specular value also controls the Frensel amount. For example, a red gloss metal sphere will reflect red reflection on parts that are facing the camera, and on the sides there's gonna be more of a white reflection. For non-metal materials specular higher than 0.5 does nothing, but metal, especially gloss one, will look more natural with the value of 1.
Yes totally! I agree! Only in this video I rather have people not touch it, because it is easy to set it wrong.
@@DamianMathew Haha, that's a proper tactics :D
In regards to IOR, I was fascinated to learn fairly recently that in reality, metals have unique IORs also. That said, in Blender I haven't noticed any difference when tweaking IORs in a metal PBR workflow. I reckon for 99% of users, using the 'correct' RGB values for metals is good enough, and saving IORs for transparent dielectrics... It just really confounds my OCD tendencies to know that my metal IORs are incorrect. It'd be cool to hear your thoughts on this topic!
Anyways cheers for the vid, it's always good to get a refresher insight on such topics, even as a self-considered 'advancing' Blender learner :-)
razeezar oh that’s interesting I didn’t know. But one thing I do know, in the PBR workflow there is no connection between metal and IOR. I used to programm PBR shaders in Unity and I never saw any connection between metal and iOR in any white paper. But also PBR is not getting reinvented, it’s pretty much based on what Pixar documented and is followed through out all softwares. So maybe there will be an update at somepoint, but probably not on Blenders side, but for example Pixar’s side and everyone will follow. But very interesting, thanks for mentioning it!
Use facing pow 2-4'ish (not important) to color mix face and edge color for artistic metals. Solid color going to white if you have no artistic color data. IOR for metals is pointless as we don't have complex IOR support yet - which is what is giving it color. These complex IORs are completely non-intuitive and pointless to adjust as user unless you get the real numbers from a site. I think Luxcore has metallic IORs if you want to play with it.
Ignore your OCD - it's not the color change from complex IOR that makes or breaks your metal appearance. It's about roughness control and maps, normal modification, aniso control with properly setup tangent (if any), and sometimes haze using separate specular lobe. Sometimes you'll need to color the metal depending on metallic grain structure. And as always, compare and adjust based on real observations.
Thank you mate
You are gold.
Thank you!
Thanks! Good explanation also for non-blender-users :) I was wondering if Fresnel for reflections is still a thing? Or do you not have to worry about that in the PBR-Workflow?
Fresnel is "Dead". Fresnel is automatically a part of the PBR workflow and is not needed on its own anymore. But I personally still use Fresnel anytime I need to break free from the Principled. For example, Velvet is a big issue but even good car paints are hard to make with the Principled. That's when I will be using Fresnel or layer weight (facing) as an alternative.
@@DamianMathew Ok good to know! Jup in the non-PBR-world it's still very alive :D
Well fuck me, I didn't know the thing about the Specular. I wonder how that will affect my future renders^^
Isn't it technically wrong to use specular maps then?
marvDE in the Physical Based Workflow, you could call it wrong, yes. But it is always good to have a control like this because often marketing shots are quite fake actually to give a cleaner look. The specular setup, kind of like a bump map (instead of normal) are legacy Workflows that still have there place, but are slowly dying. In modern shaders like in Unty3D you will not find a bump or specular input. That’s also why there is a “principled” shader, it is not a classic PBR shader but it is based on one. Also I really appreciate you watching my videos especially this was a very long one haha
@@DamianMathew x1.5 speed :)
marvDE awesome haha but yeah why not, good idea
LAST !!
real sneaky
шо за херня я подумал об этом а мне ютуб уже видосик предложил
Thank you!
Steffen Hartmann did you actually watch the whole video? I really tried to keep it short but man there is so much to say about PBR. I could’ve gone on for hours :D I am glad you like it!