How many maps for a character?

SuperFranky

Trying to find some good info about it and can't find any. So I have to ask:

How many uv maps should I use to texture a portfolio character?
What resolution is preferable today?
Is 1x4k map for the entire character is a good idea(skin on separate texture, as well as hair)? I saw someone using that for a character before...
If I don't use 1 map, how do I decide what objects to put on separate maps? How to maintain texel density then?

I hope that makes sense. Thanks in advance:)

Obscura

I'm not a character artist, but I think generally its a good idea to divide the character into different maps based on shader usage. So like one map for the hair (sort of transparency shader) one map for the skin parts (SSS) and one for the other parts (simple PBS). Also 4k still costs a lot of memory especially because you would need more than one 4k map (albedo/normal/and other shader maps such as roughness etc) so I'd rather go with multiple 2k maps.

SuperFranky

Obscura wrote: »

I'm not a character artist, but I think generally its a good idea to divide the character into different maps based on shader usage. So like one map for the hair (sort of transparency shader) one map for the skin parts (SSS) and one for the other parts (simple PBS). Also 4k still costs a lot of memory especially because you would need more than one 4k map (albedo/normal/and other shader maps such as roughness etc) so I'd rather go with multiple 2k maps.

If I'd use multiple 2k maps, how do I decide how many of these maps to use and how to separate assets better between maps?

Eric Chadwick

A couple things factor into this decision.

1. Is the character seen closeup, i.e. facial animation in cinematics. Then higher texel density for head and shoulders makes sense.

2. Do certain parts use alpha blending, while others do not? If there's hair on the head but no where else, then you don't want the whole rest of the body to have an unused alpha channel, because that eats up memory. Better to put the hair RGB and alpha in a separate smaller texture sheet.

3. Same thing goes for other textures if needed for certain effects. For example if you need a SSS map for flesh, it makes sense to limit that texture to only those UVs that need it.

4. However you can also combine maps together into different channels of the same texture, and use multiple UVs to isolate them. This decreases the # of textures to fetch from memory, but multiple UVs increases the vertex count of the model, and extra shader complexity could slow things down also.

Totally depends.

Eric Chadwick

As an example, you should check out Paul Tosca's Nyra character, which he's shared for others to pick apart.
Testing the new DX11 features of ma2013.5

rorozilla

Separating between elements that use Transparency, SSS and any other shader you where planning on using is I think your best bet like Obscura said. But personally, if you are only using one shader I wouldn't bother splitting it into different textures, Unless of course, you determined that one 4K texture was two much and one 2K was to little, then two 2K textures will be more optimal, I believe!

Hope this helps!

Eric Chadwick

For an example of a modular character design, check out Jesse Sosa's excellent example
http://oldwiki.polycount.com/SkankerzeroModularCharacterSystem

(the images aren't up on the new wiki yet)

SuperFranky

Thanks for so many answers, guys. I think I understand it now