Questions about the texel density
Hi, I've read all the information I could about the texel density, even in this forum, but still I am quite confused and with a lot of questions.
I know that it's important the texel coherency, because it has to do with the way the graphic card store the texture before it is rendered. So if something is not in that cache of 4:1, there are some texels missing.
But I don't know how much important it is in regarding the optimisation, and if I can skip sometimes that rule..
So here are my questions:
1- Do I have to keep the texel coherency along the same asset, different UVs, or also between different assets?
2- Is it important for the lightmap as well as for the textures and normal maps?
3- There are some cases where I have to stretch some UVs.
Particularly when texturing the diffuse sometimes you have to use the same texture for diferent assets in the same UV.
So sometimes is very difficult to keep that coherency, and maybe it's better to take that space in the UVs.
4- I've read something about the relation between the size of the assets and the size of the texture, but I don't understand well this.
As I believed you have to keep coherency in relation to the checker map you are using, whatever the resolution, and that's it, is not that?
I know that it's important the texel coherency, because it has to do with the way the graphic card store the texture before it is rendered. So if something is not in that cache of 4:1, there are some texels missing.
But I don't know how much important it is in regarding the optimisation, and if I can skip sometimes that rule..
So here are my questions:
1- Do I have to keep the texel coherency along the same asset, different UVs, or also between different assets?
2- Is it important for the lightmap as well as for the textures and normal maps?
3- There are some cases where I have to stretch some UVs.
Particularly when texturing the diffuse sometimes you have to use the same texture for diferent assets in the same UV.
So sometimes is very difficult to keep that coherency, and maybe it's better to take that space in the UVs.
4- I've read something about the relation between the size of the assets and the size of the texture, but I don't understand well this.
As I believed you have to keep coherency in relation to the checker map you are using, whatever the resolution, and that's it, is not that?
Replies
Basically, you want to keep things as uniform as possible. Scaling some things up or down is not a big issue, but if there are a lot of drastic changes like on the left mesh here, it will start to eat up the fill rate.
Every engine has their own ideal settings. Default Unreal3 is like 8pixels per Unreal unit I believe.
You just don't want to have a 2048x crate sitting on a 512x ground plane, ya know? It looks wrong. You want to keep things as even as possible, with few exceptions, not only for visual equality but performance.
2. Lightmap will usually be larger texels than the texture maps, because you need more unique UVs to capture the variation in lighting info. But consistent lightmap texels are generally good.
3. Yes. No set-in-stone rules here, depends on circumstances. For example, distant meshes that the player never gets close to, these can (and should have!) larger texels. Mountains, background buildings, faraway trees, etc.
However stretched UVs usually will cause the game to render a lower-resolution version of the texture (MIP), so you have to be careful about how much stretch.
4. Usually the lead artist will setup some guidelines about how many texels per game unit. Like 64 pixels per meter, or whatever makes sense for that particular game. Then everyone on the team tries to get close to this, mostly for consistency sake.
Level Designers will often scale meshes up/down to make them work in the level layout. This creates uneven density. Team needs to communicate about this when it happens, weigh the pros and cons.
Rules are meant to be broken. There always seems to be a situation that warrants it.
I knew the importance regarding the quality of the texture when the objects are closed, but I was more worried about the optimisation about the way the card process the textures.
I have another doubt, I've read that you have to align the UV island according to the pixels of the texture.. I think it's the cause of some seam issues, and it has to do again with the way the card read the texture.
But is it an important rule or it depends on some cases where the texture is small, for example? So if I'm working on a texture of 2048 I wouldn't have to be worried about that?
Also another question, what about when you are using tile textures? Is not important the texel here?
Straight pixels are better if you're doing fine lines or text. Also a bit easier to paint. But if it distorts the texture a lot, then it's kind of a tossup, not really worth it. Best way to learn is to try it and see.
Tiled textures have the same density issue as non-tiled. It generally sucks if you see small texels next to large ones. Detail textures can help hide this sometimes.
Ok so I wouldn't have to be worried about this "hardware issue", just about the quality of the texture. One problem less