For so long now, I've always thought that the traditional workflow for creating assets was you make the model first, unwrap it, then create the texture.
Though, upon coming to this forum and reading numerous threads it almost seems like the complete opposite holds true here. You make the textures first, especially tiling ones, then create the model and unwrap it.
The question and confusion I have is, how do you start with creating the textures for something if you don't have a firm idea about the final shape of the model or what the model might even be? I'd imagine if you're going off of concept art or photo reference that sort of thing becomes easier, but what if you're just creating something from scratch?
Replies
can you provide a few links where people are modeling undefined objects? and have decided to create textures first for these undefined objects?
generally you will collect materials as you make things. and build up a library. people do simple tests to develop a shader because they know they will use it. for example a grass shader. if you make grass all the time then you will probably make grass shader tests. but in all of these cases people know what they are makeing... i hope...
I think it just comes down to what you are working on, sometimes texturing first is faster, sometimes its not. All depends on the project.
You're right the previous method is the traditional way to create an asset, and by all means you can still do that, especially when working off of concept work. However there's nothing like a dark abyss between texturing and modeling.
When creating a texture you're going going to be working with a power of two sized document anyway. So regardless if it's 128/256/512/1024, the thing about these sizes is you can break them down inversely.
Such that
or
Then all you have to do is decide the sizes for your textures, generally taking into account what kind of pixel density you're looking for. Because these tile they'll just be applied with an unwrap modifier to some mesh where you can make more geometry or vertices for painting. Then you simply export to UDK. The idea behind this process is to create all the textures you'll need the first time around, so yes it does require more planning than just jumping in. But it's not so difficult to think of walls/floors/trim/misc you'll need to make an environment. Especially because when making your texture you're going to be shooting for absolute tileability meaning with or without multisuboject materials you're going to be able to mix and match to maximize distinct textures/uvs from the same source saving on memory. So really, you do know the size of everything you're creating. Especially because UDK works with Max units, and power of 2 sized objects snap together perfectly.
I think that was what you were asking?
I work in Maya so getting things sized and lined up on the grid to match UDK's takes a bit of work. Trying to think up tiling textures without concept art or photo reference though is a bit tricky, which is what I'm attempting. I'm almost starting to think the best way to start is to try and model a high-poly mesh to get a normal map that can tile on a square.
Grid in Maya is hard to set up or use, using power of 2 measurements helps ensure modular fitting (some form of numbering which is divisible by the base which would be the smallest 'unit' you have, for instance the smallest being 16 units, therefore factors of 16 and divisible similar to 16 will provide the least amount of conflict - lower the amount of measurement values really).
Creating a texture page which will represent a large portion of the world is what you want to make and this should merit the largest amount of time and energy to design out something usable and aesthetic. There's no right answer here, but the same principles of design apply - shape language, color balance, purpose vs. aesthetic.
From my own projects I've generally started by sketching on paper or mucking around zBrush - all in all, what I'm focused on is attempting to visualize the final product - however I'm not focused on efficiency or modularity yet. I'm mostly focused on making something cool and in a certain quantity (sci fi lab vs. a dungeon might call for different amounts of modular assets, in addition their forms/shapes, sizes, poly count, etc.)
Reference focus would become focused on cool elements that contribute to the final design. Concept art nowadays is also a made a lot faster, meaning the artist might of had to rely on photo reference or duplicating and skewing things - these are very helpful telltales about modular and the thought process of the world. Real world reference is similar, buildings are made from modular elements, so are most other structures, whereas vehicles are more complicated, as they should be.
Now then, getting to what I meant to say from the beginning. Dominantly in games there are 2 types of static mesh assets. There is the hero asset and the world/building assets. Hero assets obviously merit more importance, therefore it is easier to have a dedicated texture page. That means it will look specifically awesome, as the t-page is geared toward the asset.
World/building assets would be majorities of scenes, usually what the player isn't too focused on - land, walls, nature/foliage, structural components. For the most part static objects. They would benefit the most from having as little textures as they are the majority of the scene.
I'm not going to consider effects and effect meshes, such as fractured, particles, things like water, or panning materials. Those are all special case and deserve special treatment.
Okay, I'm done. Hope you get my point.
I though it's pretty normal workflow when you using only tileable textures.
Ofcourse I go back and tweak the textures, or make completely new ones, but I always try to make a basic set of textures to use.