Quote: Original post by Professor420
Ah, I think I see what you mean. How do you accomodate for the texture stretching and disproportions when texture painting? Well, its something you just do as you texture, since you always work on your texture with your model constantly updating to check your work. You also use mesh topology to guide your textures (represented in the UV snapshot), so you know where the nose is, you know where the ear is, etc. You can also use a built-in 3dpaint tool to mark landmarks like this and do rough texture layout (as opposed to using a 3dpaint tool to make your actual texture). Is that what you mean? Regardless, I think the best thing for you would be to go over a complete character tutorial- model, UV, texture, in that order, and it will give you a better idea maybe.

That's actually more along the lines of what I'm talking about. Most of what I've read about glosses over any details regarding this. I'm pretty sure the only real help is just to try it until you have enough experience, but any tips, pointers, techniques, and the like on the theory behind it is what I'm searching for.

Another fact I'd like to understand better - are these types of textures typically done entirely on the digital end? Or is it not uncommon to draw/paint a real piece, scan in and touch up?

The approach you seem to be taking is that there are painting tricks you use to get around working with deformed surfaces. Really, you should not be painting a deformed surface in the first place. Part of making a good UV map is to remove as much distortion as possible before you start painting. There is often a fundamental tradeoff between arranging your UV coordinates so that they are easy to paint on, and arranging them so that they fit together efficiently (sometimes you can squeeze in more pixels to paint on if you put things on an angle, connect them differently, etc).

For example, say you want to texture an arm. You know you will have a bracelet on the wrist that goes around in a perfect circle. It's hard to paint a line straight around the arm if the UV map has the polygons laid on an angle, or with some stretched/shrunk/etc. But if you have it mapped out in a straight cylinder, with all the polygons proportioned relative to their real 3D size (some "UV relax" functions have this option), you can just draw a straight line across and it will wrap around for you.

On the other hand, maybe you are running out of texture space and if you twist the polygons a little you can squeeze it onto the UV map so that it covers 50% more pixels. It's a choice you have to make largely based on personal preference. I've seen some people pack the entire map into square based shapes that use 100% of the texture space, but it's a real bitch to paint on.

A handy trick to checking the distortion level of your UV map is to put a texture on it that shows distortion very obviously. Here's one of my favourites. Basically, as you unwrap you just keep looking at the 3D model and see if the texture is stretching badly.

Another useful trick to use is to start out with an "ambient occlusion" map. Basically what this does is give you a starting texture with a little base lighting that can help you figure out where different shapes are.

Textures are almost always done completely digitally. However, often artists will take portions of photographs and use them as overlays to give the image more "texture" (using the word in the traditional sense, there) and detail. So real world assets do come into play, although generally not hand drawn ones.

I could see in most every case it not necessary to worry so much about direct distortion, but that all really depends on how many polys your model has. If you're running a super-low poly model (like, 2-400 polys), a lot of the shading has to be intentionally drawn into the texture, as model details (such as the sink in the eyes, or the lips pouting outward slightly) can not possibly be defined. There isn't enough polys to waste on the minute details of the model, when you have to decide whether you want a stick-figure series of appendages or texture reliant facial features. On higher polycount models (1000+), this might not be as much of a concern, as you have a few more polys to play around with on the more minute details. You can do less texture compensation this way.

However, I'm currently working on projects that deal with that 2-400 poly range, and thus, my face is basically round, with a few verticies pulled to define the nose. The rest of the body is about as well defined. Texture is now the driving force for a better quality looking model.

I do like a lot of the suggestions for being able to see a little more detail in the topology, so this may be enough to give a good jumpstart to doing this properly.