I think the problem is you've both been mis-sold and mis-informed what PBR is. PBR is literally just a set of equations that make the rendering of things follow realistic principles. It's not in-house or procedural, it's fundamental physics principles of light interaction with materials. It should be used to render scenes or games where your goal is photo realism. Now if you go and use unreal values or use the same cookie cutter tools in Quixel or whatever, then you're gonna get generic crappy looking turds. But then you look at games like GT6, Forza 7 etc, and they look natural and photorealistic. And in the end if you just want your content to look realistic then PBR is less technical and easier and faster than trying to fudge a non-PBR engine to do the same. Just look at all those horrible effects like bloom that we had about a decade ago, that 'faked' glare, but did it badly. Now we just do it as per reality and it needs no tweaking, no compromises, it just looks right automatically out of the box. At it's bare bones, PBR is just doing energy conservation and fresnel response on all reflections (basically lighting) Gloss is self-explanatory, and for dielectric materials (non-metals) you have a fixed specular value. Less gloss, softer bigger reflections. More gloss, sharper smaller reflections. Albedo is pretty much classic diffuse. AO is just the shading you'd paint into the diffuse map classically, but is now split out separately. I agree there is no 'final' art in any given texture. But really you're comparing apples and oranges. No one paints 2D textures in separate RBG channels, you just paint in all three at once in 'colour' That's exactly what a tool like Substance is trying to do for you, let you paint as if you're doing a still painting but over a 3D model, and let the software split all the information out into the various 2D channels to allow those artistic intentions be re-rendered again and again, under various lighting situations, angles, etc. At it's core, AC is a PBR engine. Now if people mis-use PBR with procedural generation systems with everything looking like it's been tumble dried with sand-paper, and/or hacked at along it's edges with a chisel, then yes it can look shite. But that's nothing to do with PBR, that's simpletons mis-using tools and doing a shitty job. But that's not exclusive to the PBR engines, or of graphics generally either. Just look at bevel buttons 'effect' on the web circa 2003. Drop-shadow buttons 'effect' circa 2011. HDR photography since about 2006 where everything is mushed and weird 'effect' Transparent jelly type buttons ala Apple Mac web style circa 2005.
I know what PBR is, it's well documented, I simply don't see a need for it when modding for AC. First person shooter or architectural presentation maybe...driving sim not so much, sure it uses RGB channels to display details here and there but that's about it. Do you have any examples of PBR textured buildings that would sit well in AC?, I can't find any good examples.
I'm fairly sure the core rendering pipeline in AC is hdr and pbr, so all buildings are already pbr shaded and 'textured' Maybe not all inputs to the shading are textured but there will be real values set. Ie, hard coded fresnel is likely 0.04 linear for perpendicular reflectivity, doing all dielectrics like car paint and glass on buildings etc. At least this is what I'm led to believe one minute, then people suggest otherwise the next. I really need to put some stuff in AC so I can see for myself but I'm pretty sure on this as I've asked someone who helped write the pbr shaders for Free Racer what AC uses, and he said it's basically the same.
I've just started to look at building textures myself, it's all a bit vague although definitely PBR at root source. One things for sure, AC's naming conventions are a bloody mess , there must be some method to their madness. I keep trying to understand these textures and my brain keeps farting.
Haha, that's what catches me out. The nomenclature reads like 2004, but the application is more like 2014. It's probably pretty simple once you get going, but there is certainly a wasted opportunity there in the shader names to convey more useful information (probably all you'd need to make the right shader choice) Dave
Sooo tempted to switch to a PBR workflow, the stone textures look awesome, still need decent photo reference tho'.
I wanted an education on PBR. Check this out> http://assettocorsamods.net/threads/pbr-texture-creation-workflow.850/
@ram, I think Megascans studio is Quixel isn't it. The Quixel tools are pretty good at taking a bitmap and turning it to normals and then AO/gloss and normals etc. Obviously there is a lot of artistic freedom in all the sliders and controls to do so. I've used them a lot for doing that kinda work. But the great thing now is that with the megascans as reference, you can see what actual real gloss and albedo (diffuse) etc values are for a material you know is like the one you want to copy, so you can tweak your own new material to be as realistic as possible. I've not done this a huge amount yet, but having real gloss/albedo references is great because 5 years ago it was scarce to have such huge data as reference. There are tons of ways to grab data though. Ie, photogrammetry can be the step up towards a full spectral scan from just a classic 2d photo because on an overcast day you can grab the normals (geometry really), and from that generate realistic AO, and being able to extract out the real albedo (or close to it), without light/shading tints from occlusion. You even see people shoot with polarised light to get specular (probably photogrammetry on a night/dark studio, with static polarised flashes) and from that derived gloss. The great thing I see in this technology area generally is that Quixel are getting Soooooo much data, their bitmap to material systems are going to have a lot of real references to look up. Ie, shoot some rock, and the algorithm will have all the megascans rocks to look at to see how the raw 2d shots of rock relate to your image of rock. I can imagine the automatic bitmap > material will become pretty good. The photogrammetry on an overcast day >> materials will become very very close to a full megascan. Scary to think that cars are now coming with lidar. The cost is going to crash. 3D scanning is going to get very accurate and cheap. The next big game/realtime graphics things are going to be neural network AI that are trained to mesh/uv these scans efficiently for realtime/game use.
Yeah Quixel, previously I was using Substance, but I find all the Allegorithmic stuff over complex and too technical. Quixel integrates with PS, It's a but clunky and funky but I've been using PS for 10 years so much prefer it's workflows. I've been "dabbling" in PBR, time to actually learn a workflow. I was watching this last night, blew me away with the detail you can get in stone textures:
I bought Quixel before the big update, and since updated to the newer ones and used them a bit too. DDO and NDO. I guess they're all just becoming one tool a bit more now. I kinda half liked the PS integration, and half didn't. It could get messy or just plain un-readable what was going on 'automagically' I did start using it mid-project though so I never really learnt all it's file handling and stuff. I'm tempted to do a sub to substance suite for a bit and have a play there too. I think having both is a good idea, though the megascans from quixel aren't cheap iirc. DDO has quite a lot of stuff in there of course, but the megascans are just awesome now. At first it was just texture based, but now they're doing meshes and LOD and everything... full PBR assets and meshes! Hmmm Dave
I was going to play with it today but a small job has landed, I'll be modelling and rendering Lime Silos all day . ..I wanna make cool stuff with Quixel Yup, I was the same, initially all the window popping absolutely did my head in, but once you appreciate what it's doing it gets easier. I'll be making all my own materials, this is a pretty good vid on the process...he records all his mistakes and gets there in the end: .
