PBR Textures in Architectural Visualisation

[reyanansari]

Hello Everyone,

I was just researching a bit about textures and Google showed me a website called "Game Textures". I was wondering that can we use their textures in Arch Viz and get the high quality realistic renders or they are just meant to be used in gaming engines?

 

 

Thanks,

Reyan

[RyderSK]

From technical point yes, textures are just rendering engine-agnostic image files. Although they often require calibration and conversion between different engines (even PBR has two workflows: Metalness/Roughness, and Specular/Glossiness, the latter being directly compatible with Vray and common raytracing engines )

 

On other hand, if you mean https://www.gametextures.com/, from "art direction" point, are not meant to be "photorealistic", but according to word of their creators, to strike middle balance between realistic and illustrated (think Warcraft, Diablo,etc..). So they're not exactly what you would use for photo-realistic architectural visualization.

[brendanrogers]

Substance Designer can create some great photo-real looking textures for use in archviz. Stone, woodfloors, brick etc, all with detailed normal and specular maps. It does take some time to create these textures so can slow down a project if your making them especially for a project and not just your library. But I recommend checking out this program, its quite brilliant.

[ansgaransgarno]

Hey there

i got an interesting problem regarding PBR texturing-this is the most visible while doing shaders for wood or fabrics,the way it should be done (according to some tutorials-viscorbels latest "reflection" one for example ) is putting the reflection on full-255 and then controlling the reflection by adjusting the reflection glossiness, the problem is (and obviously its not only in the wood and fabrics material but most annoying there) my shaders on the objects gets very white(reflecting light) approaching grazing angles-and i dont mean close to 90 degrees(which obviously should work like that) but even under like 50-60 degrees or so the texture gets almost completely washed out and white---and of course i tried to put refl.glossiness as low as 0,1 (or putting the fallof map into rglossiness slot as the tutorial said)and the result is just a fraction better(meaning texture gets little more visible/saturated/dark-if thats the goal) but it doesnt really work,sometimes putting it up to 0,8 helps but only as the light doesnt get scattered that much so the texture is more visible but thats a fake since it gets really reflective after rendering it in detail like corners and bumps-so again unrealistic---so basically the only way to make it plausible is to put reflection down to something like 10 (even that doesnt work sometimes) but then we are loosing the idea of PBR workflow. Id like to know how to deal with such problem,if u simply put the reflection almost to none or what is the work around-assuming we want to keep PBR(meaning reflection to 255 or somewhere close to that)-the main problem is it looks like this method gives more unrealistic results since its getting white-reflective way sooner than in reality plus seems like coners-grazing angles- are always white (like it should according to PBR tutorials)but after some researching it looks like thats not always the case in real world,maybe im doing this all wrong and overcomplicating stuff but since theres a chance to get closer to real world look these little questions are bugging me

thanks for the help i really appreciate it

[ansgaransgarno]

Well this is the forementioned one by viscorbel,im interested now if this is all wrong cause it seems there is some ground to this one

http://viscorbel.com/vray-materials-part-2-reflection/

[ansgaransgarno]

well this is the forementioned tutorial from viscorbel,it feels there some ground to whats he saying, all this info is very helpfull and im using it a long time already just thought this is another drop in the water to make stuff look more "photoreal"

http://viscorbel.com/vray-materials-part-2-reflection/

[RyderSK]

So they try "cheat" with a specific workflow that u get through substance or quixel etc. These programs, produce textures, not shaders. And unity/unreal can use these textures in a smart way to appear pbr.

 

rget all this PBR madness its a different thing is for engines that are not PBR to fake the refle/gloss etc.

 

Your intention is well and the rest of your advice is really good but this part is nonsense.. you're mixing too much stuff up, making up another bunch of stuff and mixing it together and just claiming it as fact.

 

Neither PBR implementation is faking anything at all regarding physical plausibility, and whether texture inputs are structured as metalness/roughness (Disney/Unreal) or specular/glossiness the result is the same. It also doesn't matter where and how they were authored. It holds true for all engines that adopted it (even Unity 3D in 5.0 iteration).

 

Hey there

Id like to know how to deal with such problem

 

On 'technical' level it's because of different implementation that masks specularity with diffuse. The game engines (like Unreal4) adopted the Disney's GGX paper in simplified way that really lets you pass on specular level and keep it full-white at grazing angle and once you go up in roughness (down in glossiness) it will show less of it.

 

In conventional ray-tracers that adopted same GGX this implementation seem to slightly differ and they don't lose specularity so much (also thanks partly to multi-bounce GI), so when you go down in roughness they seem to retain more than seems plausible and get "shiny/velvet" look. Here Thanos write correctly you need to play with Gamma value to avoid such spread (only possible in Vray, sadly solution that I have in Corona is inferior at the moment ).

 

On 'material' level no model holds perfect and truly rough materials can trap specularity to more than their microscopic surface but also thin films of dust which would further the diffusion effect. So this requires additional artistic touch and you can just lower the specularity.

 

Mind you, test your materials in scene, it's extremely important that you don't judge how they look purely based on material ball in abstract environment. Rough wood floor can look shinier than should be on material ball in Cloud-Sky HDRi setup, but then you put it into diffuse apartment on flat floor and suddenly it looks more correct.

 

80perc. of my materials keep 255 in specular slot, and I only lower it for "creative" (or "client/creative" ) reasons or when the look breaks too much.

[thanulee]

If i was talking nonsense i could use a MATERIAL str8 from substance/quixel to vray, but i cant cause substance produces MAPS not SHADERS. Ofc i can use the maps, as well as i can use textures from images, or procedurals from within vray. These maps also, are used in unity/unreal in a "smart way" to appear as realistic as possible.

 

The physical shaders are within vray, they always were there.

 

I dont know how to explain this in better english, sorry.

Edited December 29, 2015 by thanulee

[RyderSK]

If i was talking nonsense i could use a MATERIAL str8 from substance/quixel to vray, but i cant

 

Yes you can. And lot of people do. Just use your correct output calibration profile or make your own (much better). Textures are tailored to shader so you will output them in way that you can use directly, regardless of software.

 

Quixel suite has currently almost 20 profiles (from Arnold, to Vray), but you can always tailor them to absolutely any kind of shader to keep consistent 1:1 look across applications. From 3DO preview viewer in Quixel, to Vray final render.

 

Both are realistic in same way.

 

The physical shaders are within vray, they always were there.

 

No one is disputing that.

 

But Quixel Suite, Substance, and assorted game engines using PBR convention, build shaders in identical fashion. Then output varied maps to drive the shader. There is no trickery, fakery or any other seeming difference compared to offline ray-tracers.

[RyderSK]

but i cant cause substance produces MAPS not SHADERS

 

And maybe don't yell at someone who actually uses the software ;- ) Your understanding of these softwares is incorrect and limited. You constructed your own understanding and now you're preaching it as some sort of fact. What software outputs, doesn't negate what it internally builds.

 

Or, to use your logic, renderers create pixels...

[thanulee]

Maybe i m completely missing something here?

 

I am not a user of the specific software, but i dont say nonsese. I read a lot about it in order to use it or not.

Since u know it though: Can i export material settings like, 0.6 glossiness etc? Custom falloffs?? Or only textures?

And even if so... 0 advantages gain if u know how to make materials. The only thing which i ve mentioned before and u overlooked it, is that u can paint real time on a 3d object.

I dont see u having a point here, sorry

[ansgaransgarno]

Your intention is well and the rest of your advice is really good but this part is nonsense.. you're mixing too much stuff up, making up another bunch of stuff and mixing it together and just claiming it as fact.

 

Neither PBR implementation is faking anything at all regarding physical plausibility, and whether texture inputs are structured as metalness/roughness (Disney/Unreal) or specular/glossiness the result is the same. It also doesn't matter where and how they were authored. It holds true for all engines that adopted it (even Unity 3D in 5.0 iteration).

 

 

 

On 'technical' level it's because of different implementation that masks specularity with diffuse. The game engines (like Unreal4) adopted the Disney's GGX paper in simplified way that really lets you pass on specular level and keep it full-white at grazing angle and once you go up in roughness (down in glossiness) it will show less of it.

 

In conventional ray-tracers that adopted same GGX this implementation seem to slightly differ and they don't lose specularity so much (also thanks partly to multi-bounce GI), so when you go down in roughness they seem to retain more than seems plausible and get "shiny/velvet" look. Here Thanos write correctly you need to play with Gamma value to avoid such spread (only possible in Vray, sadly solution that I have in Corona is inferior at the moment ).

 

On 'material' level no model holds perfect and truly rough materials can trap specularity to more than their microscopic surface but also thin films of dust which would further the diffusion effect. So this requires additional artistic touch and you can just lower the specularity.

 

Mind you, test your materials in scene, it's extremely important that you don't judge how they look purely based on material ball in abstract environment. Rough wood floor can look shinier than should be on material ball in Cloud-Sky HDRi setup, but then you put it into diffuse apartment on flat floor and suddenly it looks more correct.

 

80perc. of my materials keep 255 in specular slot, and I only lower it for "creative" (or "client/creative" ) reasons or when the look breaks too much.

 

Hey Juraj, thanks for the long response and thorough explanation from technical and practical standpoint-very much appreciated-in a way the most valuable piece of info for me, i see there is still kind of confusion regarding these methods, but in case i go forward with the PBR way i mentioned already (reflection on 255 and controlling it by roughness/glossiness) what do i do in case when it hardly works. Thanks for advice of testing it in a scene which im doing but still this problem comes up- for example fabrics-there is just no way i can let the specularity/reflect on 255 or anywhere near that-neither 200 or 100, im forced to go low as 10 to have it roughly plausible and it still looks shiny/velvetish as u described. Same goes with trying to reproduce rough dark wood planks which are just too bright in 50-60 degree angles already,which is clearly unrealistic--the other mats looks quite allright,actually very plausible,but some of them-from time to time are just pain to make them lose the shininess

im kinda pleased someone is using this workflow as well with you saying 80perc of yours using the same method and the rest being artisticly handled according to situation, just the fact it doesnt apply generally-to all of them, makes me doubt the rest of it as well

[RyderSK]

just the fact it doesnt apply generally-to all of them, makes me doubt the rest of it as well

 

You're welcome :- ) And I absolutely understand what you mean here, it drives me mad slightly too..

 

The fact it doesn't apply generally in Vray and Corona seems to be more issue on their side, in my opinion, although Vlado@ChaosGroup seem to have different opinion (he points out the issue might be on early Disney paper adopters (from which the PBR convention originated with studios like DontNod and Epic), which might have been slightly incorrect and rectified it in this year paper, I haven't researched this further).

In Corona we have strange curve to glossiness (rather than linear line) because the devs originally thought no one will use specularity for too rough materials, which is strange...but this should be corrected in near future.

 

Fabrics ! Btw, quite topic on its own. Imho the generic GGX material model does still not apply well to something so structured as fabrics (or snow, or sand...). As fabrics are pretty complex on surface level, you basically always cheat when you create them with simple surface shader on planar geometry.

Fully diffuse model + diffuse fallof is imho still best way to go around creating very rough or velvety fabrics.

 

Regarding woods, yes, when things start to simply look wrong, it's time to break the rule, after all it's most important that things look good, not necessarily plausible. I don't think it makes the rule wrong even in case like this, I personally try to adhere to it as much as possible, it made a lot of things look much better.

[RyderSK]

Maybe i m completely missing something here?

 

I am not a user of the specific software, but i dont say nonsese. I read a lot about it in order to use it or not.

Since u know it though: Can i export material settings like, 0.6 glossiness etc? Custom falloffs?? Or only textures?

And even if so... 0 advantages gain if u know how to make materials. The only thing which i ve mentioned before and u overlooked it, is that u can paint real time on a 3d object.

I dont see u having a point here, sorry

 

Thanos, you are being obtuse.

 

I never said you're saying nonsense, I actually complimented you on good advice. I highlighted two sentences where you constructed your own very incorrect understanding, which are literally nonsense because none of that is true, one that PBR is trickery to get physical shading to game engines and other than Quixel/Substance have nothing with shader building and are useleless outside of 3D painting.

 

You you can export 0.6 glossiness, you do so through texture output, but that doesn't negate that you see real-time rendered viewport of same value with GGX BRDF, same one as in Vray.

No, you don't export custom fallof because texturing software doesn't allow you to build shading networks, but they still have internal,fully physically based shader, just without possible expanding. To build more complex network, you do in your end application, and that, can be UnrealEngine4, just as Vray. Then, you can use custom fallof to drive any feature you want. And you will again, retain identical look across application, because the convention is always physically based (even if you use slightly differing brdf, like GGX vs GTR vs Ashikmin,etc..), regardless of what inputs (values and maps) are used to drive the shader.

 

Quixel/Substance derived texture maps are not 'shader-agnostic', they're simply exported values of physically based shader stored in bitmaps, and this is what you're ignoring. They can't and shouldn't be separated, and so on abstract level, they always form a shader.

 

Regarding usefulness of texturing software suits like Quixel or Substance, they really aren't about painting in real-time only (after all, they were popular before subsets 3DO painter and SubstancePainter became available). Procedural texturing, easy baking onto texture sheets, scanned texture libraries access, batch processing,etc..

You can do the same with Photoshop ? OK. Quixel is even fully based Photoshop automated actions. Just with fancy UI and damn nice expanded overhead of features. Let's you do the same with 1/100 fraction of time. Very 'useless' if you ask me ;- )

 

Regardless, this is my last post on it, you made discussion wasted effort because you get defensive and completely stop with any reasonable reading comprehension, you just try to continue with your line, just louder. Correcting you is not attack on your ego, and I would never bother if it wasn't for your "I've read about it, and you don't need it !" 'smart' type attitude in first posts.

[brendanrogers]

You may be able to do similar things in Photoshop. But Substance Designer lets you setup a workflow so you can generate new textures easily.

this is a good example of a brick generator. I plan to have a generator for each major type of material in my archviz library. So I can load up wood floors, adjust to my clients samples and generate the normal and specular maps with ease.!

Substance designer gives awesome normalmaps, something hard to achieve in Photoshop. Nvidia plugin doesn't really give you an accurate map.

[ansgaransgarno]

And I absolutely understand what you mean here, it drives me mad slightly too..

 

The depth of level u researched it into is the proof of it, i slightly feel the same sometimes

 

Fabrics ! Btw, quite topic on its own. Imho the generic GGX material model does still not apply well to something so structured as fabrics (or snow, or sand...). As fabrics are pretty complex on surface level, you basically always cheat when you create them with simple surface shader on planar geometry.

Fully diffuse model + diffuse fallof is imho still best way to go around creating very rough or velvety fabrics.

 

 

 

Regarding woods, yes, when things start to simply look wrong, it's time to break the rule, after all it's most important that things look good, not necessarily plausible. I don't think it makes the rule wrong even in case like this, I personally try to adhere to it as much as possible, it made a lot of things look much better.

 

So basically,its completely ok to go out of the usual (specular/glossiness)way and lower it down to like 100 or 10 to get the look you are after and wont feel ashamed of failing proving the workflow 100% right, thanks for that

 

One question regarding your and veronicas last published work called "Travelling Essentials", i noticed that the materials there dont catch the "glossy edges" (due to grazing angles) that much or at all (looks truly realistic though), is it the way the lighting is done (put very smartly and created very diffuse-ly not to create those glossy white edges) in the scene or do the materials somehow differ from the "PBR", excluding the fabrics which creation youve already mentioned

 

All of this have been very educational,so thanks for that, wasnt really able to get the answers explained correctly and logically with the technical background (i should read more i guess and have to already dive into that disney paper im avoiding so long),i can now sleep tightly

[thanulee]

Regardless, this is my last post on it, you made discussion wasted effort because you get defensive and completely stop with any reasonable reading comprehension, you just try to continue with your line, just louder. Correcting you is not attack on your ego, and I would never bother if it wasn't for your "I've read about it, and you don't need it !" 'smart' type attitude in first posts.

Ya thats MY opinion man, "i read about it and i dont need it". I dont know why u find it "smart", is just my opinion which u dont seem to get ur head around.

Substance does not produce shaders, it produces textures., either u insist trying to make a point about it or not. If u cant implement the falloff/fresnel/complex fresnel which is the most important ingredient in shading, then its solid purpose is texture creation. Its a very nice program for other things but i dont get why u try persuade others that is so useful in their pipeline. Do they pay u to advertise the specific program or what?

 

My point here is that people get confused and start thinking technical while they dont have to. Developing a critical eye is much more needed rather than trying to understand terms like BRDF and diagrams with % of reflection for each material. We should think more like artists rather than technicians.

Concluding, lets not waste breath, is not rocket science: rendering is dead simple and str8 forward nowadays. Even a noob can make descent realistic renderings, is not a skill anymore is just a matter of $$$ for having gpus. And especially in arch vis where u have a single frame.

My 2 cents.