This article describes more technical things than usual about the 3D workflow for texturing models.
I wanted to write something that might help explain things I mention in future articles. I’ll keep it at an introductory level and I’m not going into too much detail at this stage.
I hope this article will be useful, I would have liked an overview of the processes involved myself when I was learning about these things. I hope to write about things in more depth in future articles as I plan to do more and learn more about texture painting myself.
Before a model can be textured it must first be retopologized and UV mapped.
Retopology is the process of building a new lower resolution mesh from the higher resolution mesh as a reference. This process is usually required if your 3D model is to be used in a game engine as fewer amounts of polygons in a mesh lower rendering times in game engines. It is also useful for meshes that are to be rigged and animated.
This can be done using manual techniques or automatic retopology techniques. Sometimes a combination of both techniques is useful. Knowledge of 3D modelling is very useful in this process for manual techniques but it is also useful to know how to fix errors and improve topology on automaticity produced meshes. Some methods of automatic retopology include tools such as Instant Meshes and Quadriflow which is now integrated into Blender.
Usually manual retopology is used on meshes intended for animation as the topology must be tailored for the rigging process to be effective on joints and moving parts. More detail is added around joints and moving parts to prevent distortion for example.
Meshes must be checked for errors and made manifold before the UV mapping process. It is also essential to fix errors in the mesh for 3D printing, but that will not be discussed in this article. I found a useful blog on this subject if you are interested in learning more about making manifold meshes. I may blog about some things I’ve learned about this as part of my 3D squirrel series of articles in the future.
Retopologized meshes are useful for a 3D artist to work with in the texturing stage. Having fewer polygons and good edge flow to work with speeds up the UV mapping process.
UV mapping is a process of projecting a 2D image to a 3D model’s surface. It is a way of making a cut out and flattened representation of your 3D model much like a sewing pattern in reverse. U and V denote the axes of the 2D image texture (X, Y, Z being the 3D space axes). There is a W coordinate perpendicular to the texture plane (but not important unless rotating).
Manual techniques and automatic techniques can be employed for UV Unwrapping and arranging UV islands for a 3D model, usually a combination of both is useful to speed the workflow.
Sometimes doing a quick automatic unwrap is fine to get started with painting, but if you unwrap it manually you will usually be able to get a better result depending on the complexity of the model. Automatic unwrapping algorithms are built into some 3D software such as 3D Coat and Blender. There is even software such as Ministry of Flat being developed that uses artificial intelligence to help 3D artists unwrap meshes. Full disclaimer I bought an individual license for this last year because I was so fascinated by it and it does work well, though with more complicated meshes some manual UV mapping is still required. I haven’t renewed the license since because I still had to manually fix things in 3D Coat. It was better than 3D Coats automatic UV mapping when I tried it with a squirrel model recently. It has probably improved since as well.
Alternatives to UV Mapping
One alternative to using UV mapping is vertex painting. This may be called different things depending on the software you use. In ZBrush it is called Polypaint. In 3D Coat it is called Vertex Paint. Vertex painting is also available in Blender.
Vertex paint information gets saved with the file. The caveat is that vertex painting needs more complex geometry to add more detail directly on the mesh and it is high memory use. However an artist is able to paint the model without needing a UV map and then transfer the texture to a UV map later on if they wish. I found this useful video if you want to learn more about it. He also has some other interesting videos on his channel.
You could also use procedural textures. Procedural textures can be generated in 3Dspace therefore don’t usually need UV mapping.
3D Sculpting, 3D modelling, retopology, UV mapping and texturing are specialisms in themselves. In big game studios I believe these processes are usually done by people working in specialist departments, but a basic knowledge of them can be useful to anyone for appreciating the process as a whole or for someone without a big team who has to do the whole process themselves. Indie game developers often learn the whole process at some point out of necessity.
From a personal perspective I learned basics of these processes from introductory courses on Udemy and by watching YouTube videos.
Because I had no prior knowledge of 3D art I thought it was useful to learn at least a little about the different stages so that I could assess what I wanted to focus on if anything.
I will also admit that I do still find the process somewhat frustrating, time consuming and technical at times but the more I learn, practice and by result demystify in my mind the less frustrating it feels. I think it good to try things because you never know if you like doing them until you do, or you might discover you don’t like doing them which is also a good way to gain more focus in the long run by process of elimination.
If sculpting is your thing and you want to get to painting your model as fast as possible but you hate 3D modelling it might be wise to try automatic retopology or ask someone else to help you with the retopology and UV mapping stage.
Since vertex painting is a possibility and if you don’t need to use your model in a game engine this needn’t stop you anyway. If you just want to paint a 3D model then you would need someone else to do the model building or sculpting things, the retopology and the UV mapping first.
You can also download free models or buy models to practice on so you don’t need to let all of that stop you if you want to practice 3D texturing and painting.
So after you have your retopologized and UV mapped your 3D model you can do all sorts of cool things if you are still with me 😉
There is lots of software you can use for texturing models, I wrote about some of it here if you need a list: Software for Painting 3D Models.
Texture can add more life and different styles to a 3D model.
Textures for 3D models are usually square ratio power of 2 numbers such as 16, 32, 64, 128, 256, 512, 1024, 2048. Usually 2K and 4K textures are common. For lower resolution models or low poly style you may only need small texture sizes. The smaller the texture size the faster the rendering times will be, but the less detail will show. For AAA games and movies they probably wouldn’t go under 2K.
Before the painting process the normals and ambient occlusion from the higher resolution mesh are usually baked onto texture images for the lower resolution meshes giving it the illusion of having the detail of the higher resolution mesh. This is useful in game engines where optimisation for rendering times is important. It is useful for the texture painting artist because it acts as a base for the painting.
I baked the normal map from a higher resolution sculpted model onto a lower resolution model for this Woodmouse 3D model I made last year, and there is a specular map for the shine on the ears, tail, feet, eyes and on the nose.
Normal maps use RGB channels for representing the direction of the 3D surface. Blue is Z height (depth), Green is how perpendicular the surface is to the Y direcion, the Red how perpendicular the surface is to the X direction. PBR textures are used a lot these days but for non photo realistic rendering another texture map that can be useful is a specular map to control areas of shine on your objects.
If you are painting or texturing your 3D object you will need to make an image for the diffuse texture or the base colours as well.
If you want to keep things simple you can use non photo realistic rendering and just use a diffuse texture like I did for this squirrel.
This can be done by hand painting the texture either in a 3D space or in an external 2D graphics program. There are other texture types to consider though.
Tiling textures or seamless textures are useful for landscapes and filling large areas with texture.
Trim textures use tiling textures to create horizontal or vertical strips of tiling patterns. Trim textures are useful in computer games for environmental assets which may share a set of themed textures. They are good for large architectural or structural objects and things like pots. Geometric border patterns are useful as well. A trim texture uses different strip widths of texture on one image- perhaps on one theme such as stone tiles to form a coherent theme. Small bands of varying texture and patterns are useful to create variation. Caveats of using trim textures are that they may produce inconsistent texture density depending on the size of the object a texture is mapped to. Advantages are that it is fast, creates coherent designs and uses less texture memory. One texture at 2K res (2048×2048) can serve multiple objects. You can also alter the same texture to create variation if you want.
PBR (physically based rendering) textures were developed by Disney to help with photorealism. Many of the modern 3D painting programs utilise PBR texture painting to aid with realism.
Here is an example of a PBR texture that simulates a metalic texture on a 3D model I made recently using art (with permission) from my sister Katherine Garner.