Texture Set Workflows: Inspired by Baddcog

[nbohr1more]

One of the things that breaks the immersion of many 3d games is the over-use of repeated tiled textures. This can be mitigated with "Texture Sets" but usually only for flat geometry.

 

Since I have come to the conclusion that this concept is a perpetual obsession of mine. I have decided to create a thread that collects my hair-brained ideas and some of the more insightful replies I've gotten about the topic.

 

Please share your thoughts or your own workflow.

 

 

Baddcog:

 

You can already do that though.

 

Just instead of making one tiling texture you have several. Alot of textures are already this way although TDM doesn't really have them.

 

Say you have your base brick (512x512), then you have the same one with dirt along the bottom, and one with drips along the top. Now you have a 'set' of vertical tiles that tile horizontally. (You can even have another with a window in it so when you go horizontally the seams match but they look different)

 

So you just need to make a set like that, but instead of being concerned with dirt/drips top and bottom you instead would want modified normal maps that have cracks running up through all 3 (of course the diffuse also needs to match). So really painting the cracks on terrain is going to give you 3 normal maps AND 3 diffuse maps (you don't want stones painted across the crack bumps... you want painted cracks through the stones to match the crack bumps).

 

So could optimize by only making the cracked texture 256 wide, and still the regular map could 'tile' up next to it.

 

You can do this using Blender, Zbrush, skulptris, Max, Photshop, Gimp, Xnormal, crazybump.... Right tool for the job.

 

I do believe someone was also working on trying to get 'megatextures' into DR though I'm not sure where that went. But basically it would be like having one huge side of a building texture instead of a combo of textures like I described.

 

From the Light Baking thread:

 

AluminumHaste:

 

This has been brought up before, even by me, but not with this particular instance. Although I have to point out again, that everything would have to be a model, which would not allow you to visportal properly. You could, I suppose separate areas with brushes and visportal that way, but those brushes would not have the baked lighting on them and would look weird juxtaposed against the nice models.

 

I downloaded and ran that guys map and yeah no visportalling is done.

 

 

 

STiFU

 

Why shouldn't it be possible to visportal it? You just have to surround the model-rooms with brushwork, so that each room is sealed and split them up with visportals intersecting with the models (No brushwork ever visible this way). You will have to make sure though that the models don't extend over multiple zones, because a model won't be rendered as soon as its origin is not within an active portal. As an example, you can have a look at NHAT 3 (forest). Goldy built the terrain with patches, which I converted to models in order to perform the vertex blending on the small path throughout the terrain (that and adding some more details to the path). The model was split into multiple models, one for each portal-zone, so that visportaling works.

 

But there is still the problem that these baked textures won't affect the lightgem, as it has been said numerous times in other threads already. So you may only use it selectively on unreachable places or in a very subtle fashion.

 

Myself in TDM Missions... :

 

....would it be possible to create "Texture Sets" groups of individual textures that when combined properly will create a complex pattern like a big patch of brickwork being exposed from behind weathered clay but across several texture tiles in a complex fashion?

 

I guess decals could do that too...

 

More decals?

 

 

From Parallax Mapped Visportal Patches:

 

Step 1: Create a basic Castle with (perhaps) some low-poly gaps representing damage

(Texture the Castle make it look as good as possible)

 

Step 2: Export the Castle into Blender

 

Step 3: Clone the Castle then modify the cloned copy into an extreme high-poly version with tons of little cracks and separated blocks/bricks making sure that some of the original texture work is still present

 

Step 4: Cut the High-Poly and Low-Poly Castles into components and create Normal Maps for the most detailed portions where things are crumbling

 

Step 6: Identify the deepest cracks in the High-Poly Castle

 

Step 7: Enlarge the Geometry of these cracks to the point where you can noclip into them when imported into Doom 3

 

Step 8: Fly into the Giant-Size cracks and capture an Environment Shot (part one of the Cube Map process)

 

Step 9: Shrink the Cube Map in Gimp (etc)

 

Step 10: Replace some of the texture work with the more detailed texture and normal map work

 

Step 11: Replace the deepest cracks with the Cube Maps (Interior Maps)

 

Step 12: Oh No! The Cube Maps look "wrong"!

 

Step 13: Stretch the proportions of the 6 environment shots so that the central "shot" is pinched a bit (lots of fiddling)

 

Step 14: After an eternity... Great crumbling castle with Cube Maps providing the illusion of deep detailed cracks

 

Models vs Brushes

Edited August 1, 2010 by nbohr1more

[nbohr1more]

So my latest thought on the matter would be a workflow like this:

 

1) Create your Prefab

2) Export as an ASE Model into Blender

3) Draw an outline around it's base

4) Flatten out (unwrap) all the UV Mapped textures into one image with the outline acting as a geometry reference

5) Modify the look and perhaps Normal Map the whole structure from a more complex version

6) Flatten out (unwrap) the new Normal Map and Texture

7) Cut it up into the original UV Mapped tile shapes using step 4 as a reference

8) Replace the Textures on your Prefab with the new ones...

Edited August 1, 2010 by nbohr1more

[Baddcog]

Don't suck me into this!

 

lol.

 

I think the point is 'you are going to do this and show US when?'

 

You've posted things from 3 threads that aren't really related.

 

As far as my post, yes there is no reason you can't take an existing texture, or make a new one and put cracks into it and/or have it tile across several tiles. The main issue is that it's probably a mission by mission thing. Just having a huge amount of textures to cover a large wall with a crack is OK for one map. But that cracks not gonna look so cool when it's in everyones map. So how useful is it?

Should we go through all the trouble for one crack? Or do we need sets of tiling textures for each material that could have a crack in it? All of a sudden one texture turns into 3-4. 20 textures turn into 100.... Lots of textures for something so specific, and when it comes down to it the geometry to make a unique crack on a map specific basis isn't going to kill performance anyway. I'd rather see and actual crack than a normal mapped 'fake' crack.

-----

 

The post about the castle is about 'baked lightmaps', aka Ambient Occlusion, which we don't use. We could use it on objects with specific textures, but again, mod bloat. Alot of our objects use 'generic' textures that are used for terrain also, so can't bake lighting onto them, and the models may not be uvmapped properly for baking anyway.

 

And the engine doesn't bake lightmaps like Hammer does (and maybe even Unreal) so while all objects might be baked the terrain wouldn't. That requires re-writing the renderer at the very least. MAYBE possible AFTER the source is released.

What that castle does is gives you a way to bake lightmaps onto the 'terrain'. Which is just a model of the terrain with greyscale lightmap textures that lays on top of the terrain and 'blends' the lightmap material with the ingame material.

Very complicated and DOUBLES the tris. Also would bloat a map quite a bit.

 

Then there's the issues with the baked shadows not hiding the player...

 

---------

The 3rd one is just about placing a '3d' image onto a visportal, so you can force it closed early (to save on rendering, processing) but still see 'the same thing through it'.

 

I have serious doubts anyone could even make this look decent once, much less expect random authors to be able to handle it.

--------

There are already several ways that map makers have found to 'hide' tiling.

 

One is having a good tiling texture. If done right they can be almost unnoticable anyway.

 

Other options are:

 

decals: put a big stain, paint, blood splatters, or even a normal mapped crack in spots that help break up the obvious tiling.

 

terrain/objects: Long boring wall, throw a couple doors/windows on it. Some pipes. A tree in front of it. Trim. Balconies...

 

Make smaller walls that don't tile as far.

 

increase texture scale so you don't get as many tiles across a surface.

 

use a wide variety of textures throughout a map.

 

use lights and shadows to highlight spots/darken spots...

[nbohr1more]

I think the point is 'you are going to do this and show US when?'

 

I'd certainly like to (best case scenario).

 

But really these are just wild speculations. I've read about various graphic techniques at beyond3d.com for years but I will admit that it has been a superficial "education" as most of the details relate to how well future hardware will render this or that. My process is to throw out these steps and find out if they are possible. If they are I keep track of them for when I actually have the skill to use them. Sometimes when I am really lucky someone like Plasticman will post how he's successfully used the "Strombine" method for one of his maps.

 

The light baking excerpts were brought up because they highlight the problems importing large models. This would be the same type of problem if you were importing large models so that walls could have (for example) large swaths of eroded plaster.

 

I was really was hoping to hear other peoples takes on what they do when they want large detailed surfaces and some of the trick's they've used. Also war stories about some difficult workflow and whether it paid-off or not.

 

Or (even better) have some of my fellow speculators come out of the wood-works and post their own crazy ideas.

 

You have such a great eye for making the most of spartan detail that it's hard to argue for methods to improve anything when you can provide such great examples that require little to no exotic methods. But I also know that the Doom 3 engine is still full of untapped potential and I am looking for those who have begun their plans to scale that mountain.