My First PBR : Will need support :)
polycounter lvl 18
Ok, so I only found out about PBR recently and have been reading up on it and watched videos. Obviously, best way to learn it is to just do it. I've just finished an asset, I started to hand paint parts of it and was thinking about going to photoreal for the purpose of PBR but I'm tempted to try it out with hand painted. First part of this process will be to work on the albedo, as you can see in my example I still have AO/shadows etc. So I'm planning on taking out all highlights and AO/shadows next. Just concentrating on the wood for now.
Replies
EDIT: And it being as stylised as it is, I think it's kinda okay to keep a little lighting information on. Kinda makes it look like a toy or smth, where someone's gone and painted it lol. Or it might even just come across as dust or wear. You could always try and push for that!
But, you could just get rid of the highlighted edges and go for a more just varied color base, like this (basically, just get rid of obvious highlights and instead try and paint in wear, and make sure any groves are dirt only, not the lighting information from the grooves):
Excuse my crappy attempt at "wear" lol, this took me all of like 10 seconds. And make sure you have an AO map pls, looks weird and it's hard to pin the stylization without it.
1) I'm using an example by Andrew Maximov who did a PBR gun in Marmoset. It's surprising to hear some of you say that gloss should just be flat colours? Here is what Andrew did for his gloss map;
2)Make sure to have AO maps? I keep reading again and again that PBR does't use AO maps (Andrew doesn't use one), I've heard in some cases some will bring it into the diffuse if it really needs it. Perhaps mine really needs it? I'll try blending it into my diffuse (or should I plug seperate AO map into AO slot in Marmo?)
Thanks again all, these are the kind of conversations I was looking for.
EDIT: And it's don't use AO maps in your Albedo... Because you have a separate AO map slot that calculates it more physically accurately! You put your AO in that slot, definitely not in your Albedo.
EDIT EDIT: I'd actually argue it is incredibly important. AO is a real life thing that happens, so in areas that will very rarely not be occluded, it looks downright bizarre if you don't have an AO map (realtime AO is lacking in finer details vs performance, that's why we bake).
http://www.polycount.com/forum/showthread.php?t=135431
Stephen even generously shared a Marmorset file you can download and study the shit out of
Ok, so my main point of reference has changed to that dagger example from Stephen. Here is where I am right now, can anyone help me understand why I'm not getting any definition? I couldn't slot my AO map in marmoset, it wouldn't let me and in Stephen's example (as I've shown at bottom of this pic) he's clearly loaded his AO map in. Also why does his AO maps show as blue? It's same with his Gloss map.
Gloss = Red channel
AO = Green channel
Cavity = Blue channel
I already had the first two and quickly converted a normal into a cavity in Photoshop just for test purposes. I slotted it into Marmoset (Thanks Bagel, didn't notice the triangles before) and well, yeah, I'm still confused. I was expecting the grooves to go dark (because they are in the cavity/AO) but everything is still flat and the same colour. I tried adding a dark tone for the grooves in the albedo and it worked but everyone says not to add too much tonal difference in the albedo so I just erased it. In terms of doing proper PBR, how do I get those grooves to go dark and how do I add tonal difference in the woodgrain?
This is what I'm aiming for (but with PBR)
Turn the Occlusion Slider up to 1, and the Diffuse Cavity to like 0.8 maybe. Might want to remove the point light for testing purposes and rely only on the IBL (maybe with the threshold tweaked) for now.
AO - this masks the ambient light contribution to the diffuse channel only. basically turning down/off the image based lighting as the AO map gets darker.
Cavity - this masks the albedo/reflective channels. each has their own slider to control the intensity of the cavity map effect, generally it looks bad when used on albedo, its primary function is to mask off reflectivity in places that the mesh and normalmaps have "conflicting information"... places like seams in cloth, or crevices in rock/wood.
So, I adjusted brightness contrast (still in xnormal) and got the following, I'm back to blue map so will use this cavity because from the two examples I've seen, they use blue gloss maps. Sound good or am I missing something here?
1) wood is very dark
2) Any ideas why my FOV goes crazy at a distance?
A previous version had FOV bound to the mousewheel, but I think we changed this in 2.05 so its ctrl+mousewheel so its harder to accidentally change the FOV. If you're still using 2.04 you should update to 2.05.
You should probably spend some time setting up nice lighting. When you're going from stylized textures with painted in directional lighting, its going to be very tough to retain the same look without putting a lot of care into your lighting. Think about what the painted in directional lighting typically represents in these sort of textures, and then try to replicate that with dynamic lights.
Honestly I wouldn't worry about packing maps into various channels and all of that stuff, just load 1 map per input for now, it will be easier to work with. Texture packing is generally something you would do when the textures are finished.
I did have 2.05 but I couldn't work out how to hide the background photo, I looked into it and all I could find were tutorials on older versions on how to do that (a sky check box that isn't in 2.05) so I just stayed with 2.04.
Interesting point about the maps? Which 1 single textures should I be keeping to for keeping things simple? Perhaps you're right, just do it one step at a time. I might just show albedo on it's own and get people to provide feedback on that without resorting to use other maps. Though it sounds like PBR is about getting multiple maps to work together so it may not be possible to go through it in steps?
Alt+RMB does the standard zoom thing, and WASD can be used for FPS style movement too.
[ame="
Actually, the sky color/blur stuff was added in 204, so you're probably back on 203 even.
[ame="
Ok, lighting: So if you were painting in a light source from the top left, try adding a point light (scene-> new light) and moving it to the top left relative to your model. Turn shadows on as well. Also play with the intensity, color, and radius(falloff) settings. You can try playing with spot lights as well.
If I want really specific lighting I will generally light the scene with dynamic lights, and then use the sky lighting as ambient light. If you want real contrasty lighting where most of your light comes from specific dynamic lights, turn the sky intensity down. If you want the sky light to fill in the shadows for a more ambient look, leave the sky intensity around the default value. Also be sure to try the different sky presets to find a lighting environment that suits your asset.
In terms of lighting, EQ's advices are spot on, you also might want to check out the tutorials on Marmoset's website.
What I'd suggest as a small, but kind of important thing is to add something to your scene that sells the scale of the door. Right now it looks a bit awkward to my eyes, because my mind can't establish the scale of things. Maybe add a person or something small that we can compare the door to.
Kepp up the good work!
Looked up the info on lighting, I'm still confused but hopefully I'll figure it out. I feel like Skylighting is what I'm after as opposed to dynamic lighting. It says the only drawback is that the lighting is baked in, not quite sure how that would be a problem for this as I only intend to use it as a static portfolio piece for now. Also, seems like skylighting gives the most 'realistic' results, not sure if that's the right word to use for what I'm after but yes I want this to look next gen in terms of materials.
So I put in a sky light in the middle (I'd deleted it before) and followed what Stephen did with his dagger by adding two similar point lights. You can see where they are by looking at the top/front views here. Also added a 1.6m cube to reference a human to help make sense of the scale.
The stone base is over bright, I think that's just because I'm not even considering them in the various textures so I probably just have them as white or something which causes this. Remember, just concentrating on wood for now.
edit : weird, in Stephen's example he doesn't have shadow casting on any of the lights. The skylight is enabled but if I toggle it there is no change to be seen at all.
The sky light is 360 hdr panoramic image, the cool thing about sky lights is that it will give you very realistic lighting, basically every point in the panorama acts as a tiny light source. The bad thing about sky lighting is that it doesn't cast shadows, simply because its much too expensive to calculate shadows from each one of those pixels which is essentially a light source. So, image based lighting tends to work really well as your ambient diffuse and specular reflection content.
Few games with use IBL along for lighting content, it is generally a mix of IBL, dynamic lights, and some sort of radiosity (usually baked but there are some other methods like voxel cone gi that can be done in real time).
Now that you've got some more interesting lighting in your scene, I think its time to focus more on what makes up your materials. To start I think you should post some reference of what sort of material you're trying to create, and by this I mean a real material not a stylized representation of material, as so far what has been shown here, the stylization is almost entirely down to the painted in lighting, which isn't really compatible with PBR.
So I would take a look at some references like:
So, if you ignore the lighting detail in these references, you can see there is a lot of interest in color variation, detail, saturation, locational effects (the doors tend to be more worn towards the bottom), all of these are potential types of detail you can add to your texture. However, Just because these are real images, doesn't mean you need to apply these concepts to your texture in a realistic manner. Your base mesh is already quite stylized, and you can add further stylization through the manner in which you paint the actual content.
Looking further at these references, its important to break down the various effects and why they occur, once you understand that, you'll know what sort of values to use in your maps.
For instance the 3rd image has multiple layers built up on the base surface of the wood, first you have the raw wood, then you have a stain, and then a lacquer and finally a coat of green paint that is mostly worn off. If you break down each of these materials you can come to logical conclusions about thier reflectance properties. The base wood (though it isn't really shown in the photo) is likely fairly bright in the albedo, with rough values in the gloss map. The stain will have darker albedo content, but similarly rough values in the gloss map. The lacquer will fill in a lot of the crevices and smooth out surface which results in a smoother value for the gloss map, and the paint will have a similar effect but probably not quite as glossy as the lacquered layer. Since none of these materials are metal they probably all have around the same reflectivity/spec value (low, about 4% linear space).
After you break down the material properties of various layers or effects, then you can think logically about where and why those effects would be applied. For instance anywhere the glossy lacquered layer has been scraped off from use would result in a rougher gloss value. Anywhere the stained wood has been stripped down to the bare wood you'll see a brighter albedo value, and on and on.
While AO and cavity maps generally should be multiplied directly over your albedo/spec/gloss, they can be very useful bases to create layer masks for specific effects, such as identifying the crevices and cavities to create a layer mask for a dirt effect, which would then use logical reflectance values for said dirt.
This is really the meat and potatoes of working a in a PBR system, first use logic and observation to determine what sort of reflectance values an element has, and then decide where and why that element would affect your asset.
Thanks again EQ, these are great notes to come back to when texturing but I'm still a bit confused as to where to begin. Let me get this right first, you say using a sky light provides good quality lighting but isn't suitable for games because it's not capable of shadows? but then at the end of that first paragraph you say it does work well? (that's if sky light = IBL which sounds like it does.)
In your second paragraph, you say stylization where edges are highlighted with a light source in the texture doesn't work with PBR. Those textures are usually what I imagine when I think stylized textures. That dagger example looked stylized yet doesn't have highlights in the albedo so you're right, I just don't yet understand what it is about that dagger that makes it stylized. Saying that it's not heavily stylized is it? I need to work out the difference between that dagger and say that pistol that is available. They sort of look the same though? Arrrgh! lol
As for providing a photo ref, that may be a good idea though I think that will end up making me to just do a more photoreal texture (with slight stylization, such as the Uncharted games). Maybe it's best to go more in that direction. Might be suicide to start PBR with heavily stylized textures, I also have less reference as not many people seem to have tutorials/examples of stylized assets using PBR.
Here are comparisons between the wood on mine and Mis-Fire's maps
Here it is in Marmo, still a long way to go
Here it is again, I lowered the intensity in my reflection/spec map and I think it's better, so I should darken the base tone in photoshop, right?
To clarify, IBL alone is typically not a good solution for lighting in games (though it can be in offline renderers which do full on ray-traced radiosity which means you get shadows from ibl). Most modern game engines do something like this:
Base IBL for the sky/clouds
Local IBL that is generated from the actual level content so the local reflections make sense, generally these IBL probes will blend to different IBL content as you move around the level. This IBL pass over-rides the base IBL pass
Ray-traced screen-space reflections, this overides the local IBL pass when the ray hits something usable.
Fully dynamic lights
Baked radiosity/global illumination
Now, not all engines use exactly these techniques or all of them, but most use at least some of them. UE4 for instance does all of the above.
Some games may only have a base IBL for the sky, this can look good especially for outdoor games, but even then you need a dynamic, directional light to represent the sun and cast shadows.
To clarify again, if your stylization is based solely on directional baked in/painted lighting, this isn't really compatible with the PBR workflow. Now, I would say that is a narrow definition of what stylized art is/can be. Pixar's movies are always a great example here, they are highly stylized, however, thier stylization has nothing to do with hand painted textures with directional lighting. Stylization can come from many different elements, including the design and shapes/details of the object, use of colors and saturation, lighting, shaders, post effects, composition, etc.
I don't think using photo reference means creating a realistic asset. The best way to understand how materials reflect light is to study real materials. Whenever you study some other artist's interpenetration of a material (a texture or a painting), you're not getting the full story, the only information you get is how that artist perceives a specific material.
Again, lets go back to Pixar. Pixar's materials are generally very realistic in nature, even if the end result isn't a photo realistic movie. So any studies you do into the physical nature of materials can be applied to stylized art content in the same manner that in can be to realistic art content. For instance, gold is typically going to reflect in the same way whether your content is realistic or stylized, and the same for skin, wood, concrete, etc.
The only real exception here is if you're trying to create a world that has completely alien materials that reflect light much differently than the materials we know of, then studying photo reference probably won't get you very far.
I think this is starting to look better. My advice would be to really push it. Right now your detailing is all very subtle. The albedo should not contain directional lighting information, but that doesn't mean it can't have a bunch of detail. I would expect to see much more variation in the color, value, and saturation of the actual wood, wood has all sorts of variation in the grain, knots, etc. The normal map content suggests a very old, beat up wood, so there is a lot of room to play with different types of wear and damage in the albedo. Dust and dirt would likely build up in the crevices, try to think about how the door would be used and what would happen in certain areas and you will start to come to conclusions as to what detail makes sense in which places.
But again I think you need to have a clearer goal of what you want the asset to look like. This can be photo reference even if the end goal is stylized, it just helps to have a target of the specific sort of wood/finish, and the detail, wear, grime/dirt that you want to represent.
TL;DR, stylize the design, stylize the object, stylize the lighting, stylize the way in which you paint the content or the specific type of detail/wear you add, but don't stylize the physical material properties (how reflective something is, how rough it is, etc), and your asset should work well with a PBR system.
Ok, I checked out reference again. I have gathered a few but this one is by far my favorite for the type of wood I'm thinking of.
I also really like the wood here too, even though it's not a photo it's clearly been well studied from photo ref,
How much would go in the roughness really depends on how the wood is finished. If its got a lacquer or clear coat, then you would get a lot of variation in the roughness depending on how fresh the coat was/how much of the coat was worn off. The refs above look like its uncoated, old, dried wood so the roughness probably would not vary a huge degree. I think more of the detail would go into the albeo (mostly color, saturation and tone detail for variances in the wood), and then also quite a lot of it in the normal map as old dried wood like this tends to have large, cracked grain.
More color varaitation would of course be added for dirt, dust, rust on the iron, etc which you can see in the gametextures door.
The top photo ref looks like the wood was originally stained a darker color but lower down the stain has worn off so it shows a ligher albedo value.
- Albedo now has more staining and variation, added highlights/shadows which despite reading as a total PBR no-no seems to improve the look a lot. So I'm not sure why some people suggest making albedos not show any light source?
- I slotted AO into Marmoset but dragging it 0-1 makes no difference at all. It has no effect on the asset.
Ok so first off, I think its important to understand the basic fundamental point of PBR, that is that all lights and materials behave in a logical and consistant manner, which means no matter how you set up your lighting, the asset will be lit as correctly as possible. Its really a two part system, part 1 is the shaders and rendering, and part 2 is having art content created with logical, calibrated values.
So, when you paint/bake in directional lighting information, you're breaking the foundation of a PBR system. Now, your shader isn't going to explode or anything, you can paint in directional light with pbr shaders, just be aware that the result you're creating isn't really physically based at that point. So if the goal is to create stylized art that is compatible with a PBR system, you should avoid baked in directional light.
Now, AO in Toolbag 2. How this works is that AO (both from baked AO and screen space AO) masks the ambient diffuse image based lighting only. There is some sound logic to this (mostly that multiplying large scale AO on top of everything doesn't make much sense, especially direct lights).
You're using dynamic lights here which in some sense will wash out the AO.
If you want occlusion to affect ambient and diffuse light from ibl and direct (dynamic) light sources, you can plug that input into the cavity map slot. Generally you don't want large scale AO, but rather micro-occlusion in the pits/pores/cracks/etc to be defined in this map. The cavity map slot in TB2 is closer to the AO slot in UE4 from what I remember.
More on the differences between ao and cavity (and all sorts of other stuff) here: http://www.marmoset.co/toolbag/learn/pbr-practice
so as of right now my asset is still PBR based? right?
That's what I was really referring to with the PBR stuff.
What I'm currently using as reference
I still like the spec method because it gives me more control when something just isn't looking right or I'm trying to make something fantastical or complex, but using Metalness as an example is a good idea-- the actual reflection content in a lot of things that aren't metal can be kinda just homogenized down to a linear 0.4 (sRGB: 0.06) value.
Recently I've been playing with that, and honestly it tends to pull things together a bit more than authoring spec in a detailed manner with values not as accurate. I'd keep away from doing that until you know exactly what you're trying to achieve by putting information like that in your spec.
Beep Boop generic "I'm learning too and could be very wrong about what to put in your spec, it wouldn't be the first time" disclaimer here.
For example, you say cracks should not be much different to the base yet I see other people's reflective maps do have tonal differences in the reflection map for indents? Even an example from the official marmoset help files you can see that? or am I misreading it? (bottom left map)
The example you have there is a camera which has parts of painted metal. The variation in the cracks that you mention is because those are areas where the paint has chipped off and is revealing metal underneath. However in your case your chipped rock just reveals more rock so it should still be the same.
In regards to the old iron looking like stone it's definitely a problem I've had too. I think part of the solution might lie in having glossier parts on edges and bits where it may have gotten worn and polished away a bit more. Some of the problem might also bee that you've gone too far in the sculpt with the chips and roughness of the surface. Not sure, just spit balling and looking at that other 3d door you referenced.
Edit: It's also really really thick, which I know is part of the stylization you wanted but it also one reasons it looks more like stone than the typical iron detailing you get on doors.
check out this reference for some ideas for further defining the iron material and edges or wear and age.
https://c1.staticflickr.com/7/6150/5922863412_29d1ec942d_z.jpg
http://www.aldgatehome.com/images/products/cira-18th-century-old-mill-slow-arch-cast-iron-window-frame-tnCs.jpg