Continuing the weekly Blender modelling practice series, here's what I worked on. This is my third week of regular modelling practice, where I am not planning to 3D-print or cast anything in silver or bronze (although this would be so cool), and my focus is entirely on improving at modelling scenes in Blender. For myself I hope focused, disciplined practice rapidly improves my abilities (which I can then use for printing and casting); for you, I hope my writing about it helps with whatever modelling challenges you may be facing, or at least gives you some interesting reading material and pretty pictures to look at for a while.

This week I've shifted from a couple of attempts at Imperial TIE Fighters, to the Rebel side of things. It seemed almost inevitable that the first Rebel fighter ought to be the obligatory X-Wing, but I decided to buck that trend for precisely that reason. Though the X-Wing is a cool model, everyone does it. And I'm sure I'll get to it in short order myself. But for now I've decided to work on the second coolest (arguably the coolest) Rebel fighter: the A-Wing!

This time I tried -- really tried -- for film accuracy. I don't think I succeeded. But I'm not really disappointed, either. It might not look exactly like Return of the Jedi, but -- in my humble opinion -- it looks at least as good as some of the video games in the franchise. I never was happy with the lighting or material properties, though. And I spent a lot of time on them. Also since I'm not tackling textures in any great detail just yet, there were some details that I just could not fully model. Maybe I lack the expertise, but it seemed like the kinds of things that should have gone into separate texture maps. So I let those go.

There is one detail that I'm particularly happy with, though. And I know it's not that big of a deal, but I really love how the engine glow came out. I'm not sure why. I just think it looks right -- more so than most of the materials, anyway.

My main focus in terms of modeling technique was on topology. I put all my effort into maintaining clean model topology and, while the TIE Fighters relied a lot of creases, here I used primarily edge loops and a few bevels. (I only really started to get the hang of it at the end of the project, though.) By the end of Week 2 I had begun to feel the pain of using creases almost everywhere. It generally worked, but it was cumbersome and occasionally led to visual artifacts. I'd already seen some YouTube videos by this point advising not to use them. They all seemed to explain pretty well what to do and how to do it, but not why. Well, I think I finally understand why.

Last week, I had to change my design a couple of times because I didn't know how to handle visual artifacts caused by the subdivision surface modifier. In hindsight, the answer was to not use creases, but to use edge loops in those areas instead. This would have greatly lowered the number of rendering artifacts, and probably would have improved performance as well, because it turns out that creases cause the subdivided vertices to cluster together, requiring significantly more polygons to produce the same look.

So this time I used edge loops and bevels. It did cause me some mental blocks trying to understand what to add and where - and I'm only just beginning to grasp how to reduce complexity in areas where it's not needed. (It's easy to understand, surprisingly difficult to actually perform.) But for the most part I think I was successful.

But although my new rule of thumb is to prefer edge loops and bevels, it begs the question: what is the exception? When do we want to use creases instead?

The answer, I think, is to use creases on areas where the edge is very small and contributes very little to the overall scene. A piece of jewelry on a character, for example. Areas where you need a hard edge, but can't justify the added labor or complexity of using more vertices. Once again, greebles come to mind. I think there's a big caveat to this, though. It seems that using creases can lead to more polygons overall, or at least, more polygons representing any particular model than would be created with bevels or edge loops. It's just that those polygons are dynamically generated by the subdivision surface modifier. So you're trading away management complexity (that is, vertices you now have to manage yourself) but increasing overall scene complexity -- and the computational power needed to render it -- in the process. This is probably fine for very simple objects, which don't have a lot of vertices to begin with and so probably don't end up generating too many in the final result. But it's not a thing to rely on. As the poly count goes up, the preference for something other than creases should go up with it.

At least, that's my understanding as of Week 3.