UV organization for new game project

Hatchiman

I’m making a few low poly models for a spec game right now and I’m just about ready to start texturing. I’ve been scouring the forum and finding a lot of good general advice, but I have several specific questions that I would like some help with. I’d just like to note that at this point, I do not have any engine restrictions with the models as I do not know which engine I’ll be using yet (though I’m leaning towards Unity). My goal is to take these models and create a video prototype for the game. Never the less, I’d like to get things done as closely as possible to the specs of what a general PC game would require, so that the video would look as close as possible to a game.

The images you see below are what I have so far, all of the textures are temporary place holder art. The camera will most likely be fixed similar to Secret of Mana, but I wouldn’t mind if the players could zoom in a bit closer to see more detail.


I’m looking for advice regarding texture size and organization of the UVS. Right now I have the textures split up as follows: there is a separate map for the head, body, eyes, mouth, eye brows, as well as a map for each of the items that the characters are wearing/holding. Each of these maps is 1k. I realize this is not optimal, I just unwrapped these quickly so that I could figure a general look.

1. Essentially, all of these characters use the same character model. The player will be able to customize his avatar with different skin colors, clothes, eye brows and mouths, gear, etc. This is the reason why I have so many separate maps.
2. Would you consider consolidating some of these maps? If so how would you handle consolidating the skin map? Right now the head is always exposed, but the hands and parts of the body will only be exposed with certain outfits. Would you recommend taking several equipment items and packing them into a single map? I thought that this could be a good way to save on memory, but if the player only has a hammer in one level and not the halberd, I might be wasting some space by having it in the same texture.
3. Right now all of the textures are 1k. I took a look at the DOTA Character Art Guide as reference and it stated that their characters use 2k textures. My characters will be about the same size, but that seems like overkill to me. Having said that, I don’t yet know what the final engine will use, so I’d like be able to work with a texture size that will look good but that I could also adapt easily to whatever engine we end up using.

Any advice and other feedback would be well appreciated.

.morph3us

well, there is no "you need to make it in this size" but if you will really never see the units more up close i would put two of the same characters, one with a small and one with a bigger texture right next to each other. if it doesnt look a lot worse the smaller filesize will be better. maybe a sharpen on the reduced texture will help to steer against some of the blurring, too.
but in the end you must compare the actual ingame size with the unit. it will never be blown up so much ingame. for reference, the warcraft 3 textures all used 512² and sometimes 256² and although it might be old, i dont think its the textures, but the renderengine, that makes this obvious

i guess the only really customizable character will be the hero, right?
maybe you could add stuff like the hellbard to the soldiers texture and then make the item "hellbard" reference the soldier texture?!
also there are a lot of baking scipts for unity, maybe that helps you with reducing draw calls in any way...

Hatchiman

Thank you xXm0RpH3usXx. I've been referencing your post over the past week as I re-lay out my new UVs.

Ace-Angel

A PC can be very forgiving, what is the actual end product you aiming at? IOS and Android?

Try stacking by item/material definition (head with heads, metal with metal, beards with beards, etc), but again, it all depends on you.
While at it, try and avoid any kind of Alpha on your model, while most upper hardware don't have issues, if you plan later porting for mobiles, those will kill your performance.

As for the what the stack will ensure when and how, think of it this, if the player has a hammer in one level, but not the next, will the characters around him have it too? Same with his gear, don't be afraid to even stack both hero and enemy assets into one texture sheet, especially if you have customization into play. Sometimes having a slightly bigger map is better then have 3 maps which get constantly loaded/unloaded on the fly.

Also, DOTA2 doesn't use 2K textures, they 'work' on 2K textures as single pieces for a single hero, before dividing them up into 256 per body part.

Also, are you guys allowed to use Vertex Colors? You could very easily drop the texture size to about 256px per certain stacks and only have specific details in them and have the Vertex color do all the more 'heavy' work on the cheap for you.

Lastly, if you can zoom on the character when he's changing gear (like in a dressing room), you most likely will have to bite the bullet and include higher rez version of the textures in their own stacks/singles, this way once the camera zooms out, you can mip-map to a lower value, but in the long run, having 1 texture with all the heads for the village while might be feasible will be a stretch for a character you can zoom on.

Cheers!