its a rock wall... i'll give you that. i'll agree that it looks like the original source image has a lot of lighting information already, which isn't good for a PBR system, since your diffuse/albedo uses most lighting info from your gloss/roughness and spec/metalness maps. it would be nice to see what your diffuse/albedo looks like to get a better idea.
well, now that I can see the image (i was too lazy to pop your link into my address bar to take a look xD) there doesn't seem to be too much lighting info. shadows are there, but there aren't a lot of in photo high-lights. nothing thats going to make it read wrong or anything of the sort. at least not that i see.
Indirect shadows are perfectly acceptable. Real-time engines are incapable of capturing small scale ambient occlusion anyway. Or rather they can't do it with proper quality anyway.
Killing shadows in photos is good idea, but it is worth remembering to do not overdo it. Especially if we are not planing on using such things as tessellation or POM to make it look more 3d.
If you kill all indirect lighting information the texture will look very flat. Which might be what you are after, but with realistic textures you still need to bake some indirect lighting information directly into diffuse texture. Even if you are zbrush zealot that sculpt everything .
@justgarry thanks, right now I'm kinda crunched on time but I really wanna make a small "beautiful corner" scene to showcase the project
@luge ya I took out most of the extreme shadows and highlights and only left the small local shadows that wouldn't be picked up in engine by the lighting
@iniside exactly! I had to leave some of those shadows so my tessellation and displacement had some back bone
I would disagree with the others - as a physically based lighting system, any shadow in your diffuse map is technically inaccurate and will only work in specific lighting conditions. If I pointed a harsh spotlight straight at this wall there would be no local shadow.
Now, it is not wrong to have absolutely no local shadows, but you still have waaaaay too much. The wall looks good on its own, but if it is put into an environment it will only look good under certain specific lighting conditions.
Here's a quick fix. This would light better under many more situations, and took me about 2 minutes to do. I think with even more tweaking you could fix the major issues.
I assure you, the wall you posted here will look terrible in any lighting conditions .
I tried it before. Though OP is free to try since he have all normal maps, and share results with us.
Ambient and indirect shadows are acceptable. Especially small scale ones, like the original photo had. Good Luck with capturing those shadows with Lightmass or even worse in real-time .
I assure you, the wall you posted here will look terrible in any lighting conditions .
I tried it before. Though OP is free to try since he have all normal maps, and share results with us.
Ambient and indirect shadows are acceptable. Especially small scale ones, like the original photo had. Good Luck with capturing those shadows with Lightmass or even worse in real-time .
Don't give bad advice, even if you couldn't get something to work well yourself, putting lighting information in the diffuse completely goes against the PBR philosophy.
Your normal map should be doing all the lighting work, and sometimes a separate AO texture.
Don't give bad advice, even if you couldn't get something to work well yourself, putting lighting information in the diffuse completely goes against the PBR philosophy.
Your normal map should be doing all the lighting work, and sometimes a separate AO texture.
This, exactly. I never said the texture I posted was a work of art, just a quick 2 minute fix to show how he can strip more lighting information out. I would agree that it looks flat and even plain ugly - surprise, your diffuse may not look as amazing when going through a PBR workflow! Your normal, height and AO (separate from its diffuse) will give you all the lighting and shadow information you need and will work in any lighting situation. Local lighting information goes against PBR, and while you can get away with it in small doses in your diffuse, you should really be applying it as a separate material pass in engine, not offline in Photoshop.
gives me more insight. I figured that having the amount of shadow information he had would be fine, not like amazing, but I couldn't think at the time as to how to level out the shadows myself. i'm still pretty new to pbr and have only done weapons through it currently sorry for giving bad advice about it.
gives me more insight. I figured that having the amount of shadow information he had would be fine, not like amazing, but I couldn't think at the time as to how to level out the shadows myself. i'm still pretty new to pbr and have only done weapons through it currently sorry for giving bad advice about it.
No worries - we all have to learn at some point. I learned when my lead yelled at me.
Indirect lighting is basically a cheat. It will look good in certain conditions where the lighting is similar to that in which the photo was taken. Any situations with a direct light source hitting the material will make it look "fake."
We're in the next gen now. You can run any texture through Knald and get a great heightmap to work off in Zbrush, and either decimate your mesh down as a module or use the heightmap in UE4 for tessellation. You'll get the right shadow detail using those methods. Lightmaps aren't perfect but your content will look consistently good in any lighting situation.
yeah, all sounds good. i'll keep that in mind when i get around to doing an enviro myself. (which will hopefully be once i get my new hard drive ) but funny how i learned something by trying to help another haha
Also if you look at the cave example content, they like to reuse the same 4 or 5 rock meshes and just rotate the meshes around. If you textures do not look good upside down, it makes then a lot less reusable and harder to cover up tiling. I know that example isn't the best for this particular texture. But imagine there was a inverted U shaped tunnel you were going to use this tiling rock texture for, one side is going to look good, the top is going to look like it's being lit from one side, and the other side is just going to look wrong.
Ok guys, instead of theoretically arguing I took texture, worked a bit on in and here are screenshots in different lighting conditions:
Does it look fake. Maybe. I leave it to you.
There is the same, where I removed more lighting, but I have added some faked AO:
Here last take, where I removed even more information:
We're in the next gen now. You can run any texture through Knald and get a great heightmap to work off in Zbrush, and either decimate your mesh down as a module or use the heightmap in UE4 for tessellation. You'll get the right shadow detail using those methods. Lightmaps aren't perfect but your content will look consistently good in any lighting situation.
I completely disagree about tessellation. It is quick road to kill performance in your game. If you start throwing it at such ordinary things like walls.
More over tessellation in UE4 is not that good in first place. It will not be able to pick as much as detail as POM from CryEngine3 (!), and it will work much slower. Especially on complex wall like this wall. On some modern bricks with a lot of space between them. Yeah it will work, though it will still not be worth be the ms you will waste on it. You either better allocate those into better characters or more complex shaders.
I will still stand that faked uniform AO is ok. It doesn't make your texture look that wrong (because it is omni directional), and it will help to bring details out.
The thing that should be absolutely removed is any directional information.
My point is that while we are in next gen, we are still no where close, to what offline rendering, and we still need to target highest FPS possible.
Inside, you are wrong, and your opinion on what you think looks better doesn't change the fact that lighting and AO baked into a albedo map screws up the math with how the shader works. If you want to bake lighting and leave harsh lighting into your albedo maps, then by all means go for it, but do not try to tell people incorrect information because you are to stubborn to learn the correct workflow.
AAA studios demand a higher quality, and that means creating better quality textures from high res sculpts or from displaced geometry, and if you are unwilling to do that than you will be quickly looked over. I had a recent art test with a studio and it was required to create the base texture and normals from a sculpt in zbrush, because that is what the job demands.
Eh? Highest FPS rendering? It doesn't matter if you use a photosource or a hand made texture, its still the same.
The fake AO is wrong, but it depends on the engine and lighting that's used.
Ass for tessellation, its EXACTLY where it should be used, on walls such as these. That's because you're more likely to see edges on these types of surfaces, and parallax doesn't really work well on edges.
As for the texture itself, it looks like a 2 minute grab from CGTextures, ran through a cheap normal map filter. Nothing more. Theres not really any work here?
Inside, you are wrong, and your opinion on what you think looks better doesn't change the fact that lighting and AO baked into a albedo map screws up the math with how the shader works. If you want to bake lighting and leave harsh lighting into your albedo maps, then by all means go for it, but do not try to tell people incorrect information because you are to stubborn to learn the correct workflow.
AAA studios demand a higher quality, and that means creating better quality textures from high res sculpts or from displaced geometry, and if you are unwilling to do that than you will be quickly looked over. I had a recent art test with a studio and it was required to create the base texture and normals from a sculpt in zbrush, because that is what the job demands.
If was this stubborn I would spend last 2h comparing different textures in different lighting conditions ;p
And my conclusion is quite simple at this point. There are cases where it looks good, and there are cases where it simply look bad. For example with only ambient light, texture look quite flat, because ambient light can't properly pick normal information, and AO can't pick information from normal map, to add some shadowing.
When you try to use directional light on some odd angles, it will certainly look faked.
I honestly don't see any golden spot here. It will be entirely up to your project what will you do.
Ass for tessellation, its EXACTLY where it should be used, on walls such as these. That's because you're more likely to see edges on these types of surfaces, and parallax doesn't really work well on edges.
Just a question. Did you tried to use tessellation on every wall in level ? And have more than 30fps ?
Last game I played where you could use tessellation on everything was Secret World, and My GTX 780 had quite the issues the render it with acceptable framerate.
For example with only ambient light, texture look quite flat, because ambient light can't properly pick normal information, and AO can't pick information from normal map, to add some shadowing.
For the tessellation... for a wall like that i would not use tessellation, if the engine supports parallax occlusion mapping i would use that for the majority of the cases.
It's true that the less light information the albedo map has the better but i kinda can see why iniside is saying what he is saying.
For the tessellation... for a wall like that i would not use tessellation, if the engine supports parallax occlusion mapping i would use that for the majority of the cases.
It's true that the less light information the albedo map has the better but i kinda can see why iniside is saying what he is saying.
Interesting read thanks!
Anyway. I have spend some more time examining, well various surfaces under different lighting conditions. If you have strong, well defined normals with quite a bit of depth, then adding any lighting information to color texture will certainly destroy it.
On other hand there are surface like moorish walls without visible mortar, and very thin spacing, or stucco plaster with very detailed patterns, where normal map, is very weak and making it to strong with do more harm than good. In that cases using some sort of faked omnidirectional detail lighting information will actually help to define surface, even if it is not the correct way.
The issue here is getting that omnidirectional information. I have tried it with cavity maps mized RG channels from normal map etc. no one true method here, bit of experimentation is need on case by case basis.
Anyway. Peace. I admit you guys have a point. But I don't agree with it to full extent.
Anyway. Peace. I admit you guys have a point. But I don't agree with it to full extent.
The problem with this line of thinking is that you have to be humble and leave your personal opinions out of the equation. In a PBR pipeline, any lighting information in the albedo / diffuse is technically incorrect. You can't argue or disagree - it no longer is about artistic style but cold, hard shader math. If you are using a non-physically based pipeline, feel free to bake your AO, cavity, emissive, and who cares what else into a diffuse.
As an artist, it is difficult to cope with this. Our goal is to make something look good, often under particular lighting conditions, with specific color schemes and a composition to take into account. But PBR makes all of us technical artists, as opposed to just artists. Understand the math and make it work for you, instead of trying to work against it, and you'll find your work will look better for it. Falling back on AO is a cheap solution and avoids the larger problem.
Saw this on Facebook. The fact that PBR is so confusing to so many people I may have to complete my doc I was doing about PBR and texture process below is a copy pasted segment with some more added info. Honestly PBR is not that hard of a concept and seems like everyone is just throwing in a lot of opinions without knowing what it means. You can still have artistic liberties, but I would have to say the steps taken here are really wrong. Your lighting conditions show this even more.
Steps to take depending on workflow...
Zbrush (best workflow)
-Model bricks alike Tate tutorial http://pixologic.com/zclassroom/homeroom/lesson/environments-with-tate-mosesian/
-Make your albedo texture either in Zbrush by poly painting
-bake your passes in zbrush or x-normal
-Make your passes Albedo, Roughness... etc
-set up shader and done.
Ndo, Substance, Knald
-Make your normal and height by extracting information from photo ref.
-Make sure to stack overlay different normals to get best results
-Make your passes Albedo, Roughness... etc
-setup shader and done
Crazybump (worst) <- I know a lot of people use it... I really think Crazybump is crutch for artists. -Make your normal and height by extracting information from photo ref.
-Make sure to stack overlay different normals to get best results
-Make your passes Albedo, Roughness... etc
-setup shader and done
{{{{{{This is a copy and paste from previous thread I posted on}}}}}}}
Macbeth Chart is good for referencing colors and or balancing photos.
Here is a shot of the passes. The guys that made Remember Me admit that too much micro detailing was added to the Albedo it was a learning process for them too, so you still get very dark textures with some shadow information but in general all textures are pretty clean. In general I would go cleaner than some of these below to be safe and add more Albedo color information if needed later.
How to Author Textures Workflow
Normal and Roughness
The normal is usually authored in Zbrush, NDO, Crazybump, Substance Designer, or many more. Roughness usually authored when creating your material masks either in Photoshop or my case Substance Designer using real life reference first and also trying to stick close to the real life measurements.
a. Normals - represents the surface forms and many cases the larger roughness scale of the objects.
b. Roughness - Roughness is really how smooth or rough a material is, it is usually described by micro facets on the surface the more rough the less reflective and less rough more reflective.
c. by combining these two in harmony the Diffuse and other passes will just work.
(From my own experience making normals that have too much micro detail will only hurt your end result also be aware that both normal and roughness help in creating the illusion of a material.)
Albedo
Albedo is your color or diffuse map. The only big difference for game artists is that we no longer represent shadow or AO information baked into the Albedo. Usually authored in photoshop, Substance Designer, or any 3D painting program.
a. No AO or very little AO baked into maps. (stick with no AO to be safe) AO can be used as dirt gathering think about materials happening and history of the texture when thinking of using AO pass use it as dirt this is where you can have artist choices.
b. Color values should be checked with real life measurements using the Macbeth chart as a guide.
C. You can and should have freedom to control colors if it fits the style of your game this is not barrier... Charts and Data do not make balanced compositions... the artist eye does.
All other maps are Engine specific Those are authored in any paint program usually Photoshop.
I will stress though the maps you do not want to screw around with is Roughness and Normals. These two will give you the best material response in the end. Albedo just has to be a good enough close to the color of the material as possible and all other maps are important just from my experience so far Roughness and Normals are far more important everything else just kind of drops in to place as you do the process of creating the textures.
So to be short and clear. IMO... of course.
Start like this...
-Create your Normal (Hopefully sculpted or baked from higher poly for best results)
-Bake out passes like Curvature, AO, Height, and IDs.
-Make your Albedo using Curvature, AO, and Height using balanced values for the material being made.
1. AO - Can be used to add dirt in crevices not as a lighting technique.
2. Curvature - Same you can use it with level adjustments in Photoshop for the curvature to get your peaks and valleys to get wear and dirt masks.
3. Using the passes for your albedo will start to create masks for you which you can use for Roughness and other passes used for BRDF.
-Create your roughness passes with masks baked out or made.
-Set your Albedo, Roughness, Normal, Specular, Metallic, and AO into your shader.
Just remember Roughness and Normal are very important and others can become very important depending on the material you are trying to mimic especially when you go into metals and such.
I'd say, as a for fun artist that's also a graphics programmer, that it's not even totally necessary to use the Macbeth chart. Heck mostly I'd say you use that while taking reference photos yourself so you can check colors and help eliminate environmental lighting. You can still do handpainted stuff with PBR.
But definitely no AO baked. It's a dream to be able to get everything "working right" and not fall back on hacks like baked in AO. Using UE4 on the high (or even mid) end will have any game or production absolutely require you not to bake AO at all, unless it's some very very specific and odd art direction.
The entire idea is to model everything in a kind of perfect isolation. Each texture and mesh and etc. to be imagined as just perfectly by itself, in some otherwise infinite void. Then the idea is to go and use the actual game environment to fill in all the environmental stuff as much as possible. Ideally this should make all textures and meshes perfectly reusable everywhere, without any unexpected outcomes no matter what sort of lighting or etc. you stick them in.
Oh yeah the Macbeth chart is good for photo ref I agree macbeth chart is hardly used but I think it is good to know about it. Also you can still have painted textures work with PBR. Think Frozen or Pixar movies.
WOW! Thank you guys for all these awesome resources! I know its not perfect, but this was my first attempt at creating an asset using the PBR pipeline.
I really appreciate all the paint overs, examples, and charts you guys made to demonstrate your points, especially @MooseCommander, @rogelio and @iniside.
Unfortunately, this project was a surprise to me and only had a few days to complete it. And it looks like an obvious tiling texture because my prompt was to generate a tiling texture straight from a photograph. The normals and other maps were generated from zbrush using masks. They all need some serious work.
I plan to revisit this asset after graduating when I have a few moments to relax and really learn how to correctly create some awesome looking PBR materials. Again I am extremely thankful for all the responses!
This place is full of arseholes. Each of them having their own take on things.
The thing is?
Each of these arseholes has a point.
I learnt the hard way, ignoring the advice of the few guys here who could improve my work. It took me too long to come around to it.
Fact of the matter is theres guys and girls here who will kick your arse. Even after 10 years more of trying. Listen to them. Learn from them. Don't be so quick to call everybody wrong because its not what you're used to. I did it once... In hindsight, I couldn't have been more wrong.
Pretty much everything I know today I owe to this place. Use it wisely.
Replies
Post image like this without spaces
[ IMG ] http://i.imgur.com/iHnh3RQ.jpg [ /IMG ]
[ IMG ] http://i.imgur.com/yoQ64sV.jpg [ /IMG ]
You should probably post the base color image as well, I feel like there is probably too much lighting information in it for a PBR based engine.
The prompt is to generate a material from photosource
Are you planning to build a scene using this?
Killing shadows in photos is good idea, but it is worth remembering to do not overdo it. Especially if we are not planing on using such things as tessellation or POM to make it look more 3d.
If you kill all indirect lighting information the texture will look very flat. Which might be what you are after, but with realistic textures you still need to bake some indirect lighting information directly into diffuse texture. Even if you are zbrush zealot that sculpt everything
@luge ya I took out most of the extreme shadows and highlights and only left the small local shadows that wouldn't be picked up in engine by the lighting
@iniside exactly! I had to leave some of those shadows so my tessellation and displacement had some back bone
Now, it is not wrong to have absolutely no local shadows, but you still have waaaaay too much. The wall looks good on its own, but if it is put into an environment it will only look good under certain specific lighting conditions.
Here's a quick fix. This would light better under many more situations, and took me about 2 minutes to do. I think with even more tweaking you could fix the major issues.
I tried it before. Though OP is free to try since he have all normal maps, and share results with us.
Ambient and indirect shadows are acceptable. Especially small scale ones, like the original photo had. Good Luck with capturing those shadows with Lightmass or even worse in real-time
Don't give bad advice, even if you couldn't get something to work well yourself, putting lighting information in the diffuse completely goes against the PBR philosophy.
Your normal map should be doing all the lighting work, and sometimes a separate AO texture.
This, exactly. I never said the texture I posted was a work of art, just a quick 2 minute fix to show how he can strip more lighting information out. I would agree that it looks flat and even plain ugly - surprise, your diffuse may not look as amazing when going through a PBR workflow! Your normal, height and AO (separate from its diffuse) will give you all the lighting and shadow information you need and will work in any lighting situation. Local lighting information goes against PBR, and while you can get away with it in small doses in your diffuse, you should really be applying it as a separate material pass in engine, not offline in Photoshop.
No worries - we all have to learn at some point. I learned when my lead yelled at me.
Indirect lighting is basically a cheat. It will look good in certain conditions where the lighting is similar to that in which the photo was taken. Any situations with a direct light source hitting the material will make it look "fake."
We're in the next gen now. You can run any texture through Knald and get a great heightmap to work off in Zbrush, and either decimate your mesh down as a module or use the heightmap in UE4 for tessellation. You'll get the right shadow detail using those methods. Lightmaps aren't perfect but your content will look consistently good in any lighting situation.
Does it look fake. Maybe. I leave it to you.
There is the same, where I removed more lighting, but I have added some faked AO:
Here last take, where I removed even more information:
More over tessellation in UE4 is not that good in first place. It will not be able to pick as much as detail as POM from CryEngine3 (!), and it will work much slower. Especially on complex wall like this wall. On some modern bricks with a lot of space between them. Yeah it will work, though it will still not be worth be the ms you will waste on it. You either better allocate those into better characters or more complex shaders.
I will still stand that faked uniform AO is ok. It doesn't make your texture look that wrong (because it is omni directional), and it will help to bring details out.
The thing that should be absolutely removed is any directional information.
My point is that while we are in next gen, we are still no where close, to what offline rendering, and we still need to target highest FPS possible.
AAA studios demand a higher quality, and that means creating better quality textures from high res sculpts or from displaced geometry, and if you are unwilling to do that than you will be quickly looked over. I had a recent art test with a studio and it was required to create the base texture and normals from a sculpt in zbrush, because that is what the job demands.
The fake AO is wrong, but it depends on the engine and lighting that's used.
Ass for tessellation, its EXACTLY where it should be used, on walls such as these. That's because you're more likely to see edges on these types of surfaces, and parallax doesn't really work well on edges.
As for the texture itself, it looks like a 2 minute grab from CGTextures, ran through a cheap normal map filter. Nothing more. Theres not really any work here?
And my conclusion is quite simple at this point. There are cases where it looks good, and there are cases where it simply look bad. For example with only ambient light, texture look quite flat, because ambient light can't properly pick normal information, and AO can't pick information from normal map, to add some shadowing.
When you try to use directional light on some odd angles, it will certainly look faked.
I honestly don't see any golden spot here. It will be entirely up to your project what will you do.
Just a question. Did you tried to use tessellation on every wall in level ? And have more than 30fps ?
Last game I played where you could use tessellation on everything was Secret World, and My GTX 780 had quite the issues the render it with acceptable framerate.
http://miciwan.com/SIGGRAPH2013/Lighting%20Technology%20of%20The%20Last%20Of%20Us.pdf
It really depends on the engine and ad hoc implementation.
For the tessellation... for a wall like that i would not use tessellation, if the engine supports parallax occlusion mapping i would use that for the majority of the cases.
It's true that the less light information the albedo map has the better but i kinda can see why iniside is saying what he is saying.
Interesting read thanks!
Anyway. I have spend some more time examining, well various surfaces under different lighting conditions. If you have strong, well defined normals with quite a bit of depth, then adding any lighting information to color texture will certainly destroy it.
On other hand there are surface like moorish walls without visible mortar, and very thin spacing, or stucco plaster with very detailed patterns, where normal map, is very weak and making it to strong with do more harm than good. In that cases using some sort of faked omnidirectional detail lighting information will actually help to define surface, even if it is not the correct way.
The issue here is getting that omnidirectional information. I have tried it with cavity maps mized RG channels from normal map etc. no one true method here, bit of experimentation is need on case by case basis.
Anyway. Peace. I admit you guys have a point. But I don't agree with it to full extent.
The problem with this line of thinking is that you have to be humble and leave your personal opinions out of the equation. In a PBR pipeline, any lighting information in the albedo / diffuse is technically incorrect. You can't argue or disagree - it no longer is about artistic style but cold, hard shader math. If you are using a non-physically based pipeline, feel free to bake your AO, cavity, emissive, and who cares what else into a diffuse.
As an artist, it is difficult to cope with this. Our goal is to make something look good, often under particular lighting conditions, with specific color schemes and a composition to take into account. But PBR makes all of us technical artists, as opposed to just artists. Understand the math and make it work for you, instead of trying to work against it, and you'll find your work will look better for it. Falling back on AO is a cheap solution and avoids the larger problem.
Saw this on Facebook. The fact that PBR is so confusing to so many people I may have to complete my doc I was doing about PBR and texture process below is a copy pasted segment with some more added info. Honestly PBR is not that hard of a concept and seems like everyone is just throwing in a lot of opinions without knowing what it means. You can still have artistic liberties, but I would have to say the steps taken here are really wrong. Your lighting conditions show this even more.
Steps to take depending on workflow...
Zbrush (best workflow)
-Model bricks alike Tate tutorial http://pixologic.com/zclassroom/homeroom/lesson/environments-with-tate-mosesian/
-Make your albedo texture either in Zbrush by poly painting
-bake your passes in zbrush or x-normal
-Make your passes Albedo, Roughness... etc
-set up shader and done.
Ndo, Substance, Knald
-Make your normal and height by extracting information from photo ref.
-Make sure to stack overlay different normals to get best results
-Make your passes Albedo, Roughness... etc
-setup shader and done
Crazybump (worst) <- I know a lot of people use it... I really think Crazybump is crutch for artists.
-Make your normal and height by extracting information from photo ref.
-Make sure to stack overlay different normals to get best results
-Make your passes Albedo, Roughness... etc
-setup shader and done
{{{{{{This is a copy and paste from previous thread I posted on}}}}}}}
Macbeth Chart is good for referencing colors and or balancing photos.
Macbeth Chart
"Remember Me" has a great overview of PBR a lot can be gleamed from this.
PBR workflow for Remember Me
Here is a shot of the passes. The guys that made Remember Me admit that too much micro detailing was added to the Albedo it was a learning process for them too, so you still get very dark textures with some shadow information but in general all textures are pretty clean. In general I would go cleaner than some of these below to be safe and add more Albedo color information if needed later.
How to Author Textures Workflow
Normal and Roughness
The normal is usually authored in Zbrush, NDO, Crazybump, Substance Designer, or many more. Roughness usually authored when creating your material masks either in Photoshop or my case Substance Designer using real life reference first and also trying to stick close to the real life measurements.
a. Normals - represents the surface forms and many cases the larger roughness scale of the objects.
b. Roughness - Roughness is really how smooth or rough a material is, it is usually described by micro facets on the surface the more rough the less reflective and less rough more reflective.
c. by combining these two in harmony the Diffuse and other passes will just work.
(From my own experience making normals that have too much micro detail will only hurt your end result also be aware that both normal and roughness help in creating the illusion of a material.)
Albedo
Albedo is your color or diffuse map. The only big difference for game artists is that we no longer represent shadow or AO information baked into the Albedo. Usually authored in photoshop, Substance Designer, or any 3D painting program.
a. No AO or very little AO baked into maps. (stick with no AO to be safe) AO can be used as dirt gathering think about materials happening and history of the texture when thinking of using AO pass use it as dirt this is where you can have artist choices.
b. Color values should be checked with real life measurements using the Macbeth chart as a guide.
C. You can and should have freedom to control colors if it fits the style of your game this is not barrier... Charts and Data do not make balanced compositions... the artist eye does.
All other maps are Engine specific
Those are authored in any paint program usually Photoshop.
I will stress though the maps you do not want to screw around with is Roughness and Normals. These two will give you the best material response in the end. Albedo just has to be a good enough close to the color of the material as possible and all other maps are important just from my experience so far Roughness and Normals are far more important everything else just kind of drops in to place as you do the process of creating the textures.
So to be short and clear. IMO... of course.
Start like this...
-Create your Normal (Hopefully sculpted or baked from higher poly for best results)
-Bake out passes like Curvature, AO, Height, and IDs.
-Make your Albedo using Curvature, AO, and Height using balanced values for the material being made.
1. AO - Can be used to add dirt in crevices not as a lighting technique.
2. Curvature - Same you can use it with level adjustments in Photoshop for the curvature to get your peaks and valleys to get wear and dirt masks.
3. Using the passes for your albedo will start to create masks for you which you can use for Roughness and other passes used for BRDF.
-Create your roughness passes with masks baked out or made.
-Set your Albedo, Roughness, Normal, Specular, Metallic, and AO into your shader.
Just remember Roughness and Normal are very important and others can become very important depending on the material you are trying to mimic especially when you go into metals and such.
http://youtu.be/DMO4X9leTG8?t=3m21s
And this is the results at render time, http://cdn1-www.playstationlifestyle.net/assets/uploads/2013/06/killzoneshadowfalle3screenshot2.jpg.
I'd say, as a for fun artist that's also a graphics programmer, that it's not even totally necessary to use the Macbeth chart. Heck mostly I'd say you use that while taking reference photos yourself so you can check colors and help eliminate environmental lighting. You can still do handpainted stuff with PBR.
But definitely no AO baked. It's a dream to be able to get everything "working right" and not fall back on hacks like baked in AO. Using UE4 on the high (or even mid) end will have any game or production absolutely require you not to bake AO at all, unless it's some very very specific and odd art direction.
The entire idea is to model everything in a kind of perfect isolation. Each texture and mesh and etc. to be imagined as just perfectly by itself, in some otherwise infinite void. Then the idea is to go and use the actual game environment to fill in all the environmental stuff as much as possible. Ideally this should make all textures and meshes perfectly reusable everywhere, without any unexpected outcomes no matter what sort of lighting or etc. you stick them in.
I really appreciate all the paint overs, examples, and charts you guys made to demonstrate your points, especially @MooseCommander, @rogelio and @iniside.
Unfortunately, this project was a surprise to me and only had a few days to complete it. And it looks like an obvious tiling texture because my prompt was to generate a tiling texture straight from a photograph. The normals and other maps were generated from zbrush using masks. They all need some serious work.
I plan to revisit this asset after graduating when I have a few moments to relax and really learn how to correctly create some awesome looking PBR materials. Again I am extremely thankful for all the responses!
The thing is?
Each of these arseholes has a point.
I learnt the hard way, ignoring the advice of the few guys here who could improve my work. It took me too long to come around to it.
Fact of the matter is theres guys and girls here who will kick your arse. Even after 10 years more of trying. Listen to them. Learn from them. Don't be so quick to call everybody wrong because its not what you're used to. I did it once... In hindsight, I couldn't have been more wrong.
Pretty much everything I know today I owe to this place. Use it wisely.