Modular assets or texture blending.
Hi there. I am in trying to get my head around the best approach to take for creating a natural environment assets for my game. The main problem i am having is when to use unique unwrapped meshes and textures and when to reuse and tile textures. I am aware that having unique maps for many assets would cause the project to balloon in size but if they are used in a modular fashion then it may be the best way to go. Plus modern games are very large anyway. I have included an image from ratchet and clank as an example. The rocks in the distance in the image, do you think they use one big map to create all that nice detail or have they tiled a base texture and then overlaid the details in the using the shader(texture blending)? Or could they be clusters of smaller unique rock assets? Any thoughts would be much appreciated, thanks.
Replies
Large areas usually use tiled textures with blending. And closeup meshes often use atlased/baked textures. But you can mix these techniques together.
It's important to keep the texels at similar sizes across different meshes, otherwise some parts will look more blurry than others, which tends to look weird to the player. So be careful with texel density.
Thanks for the input. I don't suppose you would care to fathom a guess as to how the rocks in the distance(of the pic) have been produced? They look to me like they might have been created by pulling together multiple smaller rock modules to create the larger face.and then perhaps a grass material vertex painted over the top to create a more unique model?
Can you please draw a line around the area you're asking about?
If it's under the palm tree on the left side, that's looks like a tiled grass texture blended with a tiled rock texture, then some uniquely-mapped rocks shoved into it.
We have some tutorials and examples here
http://wiki.polycount.com/wiki/EnvironmentSculpting
http://wiki.polycount.com/wiki/MultiTexture
http://wiki.polycount.com/wiki/Terrain_in_The_Last_Of_Us
https://polycount.com/discussion/159954/the-environment-art-of-halo-4/p1
Yes that would have been a good idea in the first place, here you go:
ps. thanks for the links i will check them out.
Those are meshes smashed into each other. You can see they didn't rotate the ones on the bottom, an obvious repeat.
The green tops up above could be a couple things... A shader that blends in a color or texture onto surfaces facing up. Or, it could be blending based on painted values, like vertex color.
My guess is that each rock asset has 1-2 unique maps; Normal and a packed mask map of some kind - there's definitely a curvature map in there. Then there's a few tri-planar mapped textures blended in and/or some vertex painting.
Hi all.
I was googling when it is better to use tileable texture blending and when it is better to create the textures with photoshop or Substance Painter for models for film and video games, I came across this forum topic, I couldn't resist the temptation to ask since I can't find anything that convinces me on Google.
Considering that mixing 1k or 2k tileable textures from color mask layers consumes less resources and the results are sharper than a single 4K texture created in substance, I was wondering why this technique is not used more in smaller asset models?
Wouldn't it be a good idea to use color masks to blend 2 or more textures on complex models like dirty cars, chairs, tables, blinds, etc.
Is it a good idea to use this technique on detailed decoration models for video games and movies?
What limitations could I find?
With an optimized software to play with the mixture of tileable textures and masks, couldn't I get results like in substance painter without the limitation of the texture pixels and consume less system resources?
Thank you very much for your attention
@orangesky
Its used quite a bit in AAA. See Cyberpunk 2077 for liberal use of this method.
It can consume less VRAM, but you trade that for greater material complexity.
Thank you very much for your answer.
I found this wonderful interview where they talk about it.
https://substance3d.adobe.com/magazine/cyberpunk-2077-a-world-full-of-substance/
It would be great if artists who are not part of CD Project could also have the opportunity to use the method in an optimized way for substance painter.
Do you know of the existence of any script or software that can be used by any artist?
Substance isn't required to do this - what you need is a game engine that supports blending using shaders. They're just using substance to create masks. You could make these masks in photoshop or any image editor. It's just easier to do it using substance painter.
Both Unreal and Unity could do this. Unreal Engine would be the most obvious choice for me since it has a layered shader system already set up:
https://docs.unrealengine.com/4.26/en-US/RenderingAndGraphics/Materials/HowTo/CreatingLayeredMaterials/
Many games use a similar technique. It's not unusual.
in this scene I used layered shaders but used vertex color as a mask instead of a texture.
https://youtu.be/T8ZqoaS6OsI
you can see the vertex color masks at 0:03 in this video:
https://youtu.be/HXD6qF5hysg
Thank you for your contribution!
I see that the technique is not as unusual as it seems, I am happy to see it being used nowadays, the main problem was that I didn't know how to do it in Substance, I thought that this software was not prepared for it but I have seen that it is and I am very happy about it.
Artists in general don't seem to use this technique very often and I don't understand why it is not used more
Now I just need to know in which cases it is better to use the layered material technique and in which cases it is better to flatten the details of a model in a single texture.
According to your criteria, what limitations, advantages and disadvantages would have the technique against conventional texturing?
Congratulations for your work, I love the result of Throne Rooom, I really like the aesthetics, it reminds me of the Fortress map from Rainbow Six Siege.
Depends on the technical restraints of the game and on the art direction.
If assets need to fit a specific specialized style, and need to have very few draw calls, then baking them into unique sheets might make the most sense.
Another example using modular design and layered shaders together to excellent effect:
https://polycount.com/discussion/227450/the-ascent-damn-its-pretty
See Shubham Kumar's tutorial here, using Substance Painter to make masks to blend trim sheets
http://wiki.polycount.com/wiki/Modular_environments
Another great example is The Matrix Awakens demo that Epic recently made. They go over the details of the texturing in this video starting at 28:00
https://youtu.be/xLVJP-o0g28?t=1722
@orangesky
"Artists in general don't seem to use this technique very often and I don't understand why it is not used more"
Its a lot easier to make pretty portfolio pieces with unique textures, and setting up this kind of system is more of a tech art thing :)