I had written about “creating a column” in Blender. I posted a few in-progress screen shots of the “Starlyte” Lazer Tag blaster that I thought I might try modeling. Neither of these came out anywhere near good enough to work with, so I opted to fall back on an “original design” — a sci-fi style wall piece that could be used in a game, maybe, or as a tile in a 3D VTTRPG application.
Despite having a somewhat coherent image in mind, I simply could not get this damn thing to work.
This is a new model based on my original design. It started out as a cube, and I employed BoxCutter — the single most useful add-on for Blender, hands down — to carve out the top and bottom extrusions, and the horizontal and vertical center channels. I also used it for the top vents, the divot on the left side, the borders of the various panels (which are basically insets), and the lozenge-shaped whatevers on the right.
What is BoxCutter? It’s a non-destructive way of carving out a hard-surface design. “Destructive” modeling would require me to alter the original cube, which would eventually preclude me from being able to CTRL+Z my way out of an unmitigated design disaster. Using boolean modifiers — shapes that can add or remove chunks from other shapes — I can remove designs at any time I want, one by one. Using these modifiers makes it easy to revert the above design back to the original cube by just jettisoning all of the boolean cuts that BoxCutter allowed me to make.
One issue is that eventually, those booleans need to be applied in order to move on with the baking phase. Once I did this, my model changed significantly.
Now you can see the geometry that the booleans have wrought. This isn’t so bad, really, although I now realize I should have probably broken up those large face islands into smaller bits, and then triangulated everything. Since this is the “high poly” version, having a lot of geometry is OK; it will serve as a template for “baking down” various image files like normal maps, ambient occlusion, and material IDs.
Most of the issues I ended up with were only found during the baking. Doing it the long way in Blender requires the creation of different images onto which the high poly details are “baked” in a way that they fit onto the unwrapped UV of the low poly model. Think of it as peeling off the detailed skin of the high poly version, complete with all of the indents and insets and shadowy crevasses, and laying that skin on top of a shape that kinda looks like the one you just peeled, at least in profile. The bakes tell the renderer how the light should bounce, so when applied to a model with no actual geometry to do that bouncing itself, we still get a really believable model complete with details.
Well, sort of. Generally, I am pleased with the model itself. I’m not ecstatic, and that makes me incredibly sad and frustrated. Coming up with a design and details is very hard for me. Whether that’s old age or just the result of a complete cerebral break-down, nothing creative seems to be working for me as of late. I’ve learned that getting the details just right is difficult for many people. This is why I’m pleased with the wall design, but I’m nowhere near happy with it.
Part of it has to do with the frustration of texturing. This example above is a real flippin’ mess, courtesy of Quixel Mixer. My brain got trained on Substance Painter and I was recently watching a video in which Painter was used. I immediately understood everything being done in the video where Painter was involved. When I tried to apply similar techniques in Mixer, I was at a loss; I’m not saying that I’m an expert at Painter, but something about Mixer ultimately feels…off to me. So while I’m happy that Mixer is free, I think it’s still a less-developed product compared to Painter, and I am considering sucking it up and taking on that ludicrous monthly fee. Screw you, Adobe.
I can’t completely blame the tools, though (for the record, I am also a tool in this situation). I found some baking issues (?) once I got the model into Mixer.
The upper vents are normal-baked and aren’t actually “3D” in this model. The divot on the left is (as it follows the rule that anything that gives the model its silhouette should be modeled, and anything that gives the model detail should be baked). Something is weird with the bake on the vents and also around the divot; I blame the Ambient Occlusion map.
There are also similar issues on the other side of the model.
This one would probably be hard to see if I weren’t actively looking for it. Whatever is going on with the issues on the left is apparently plaguing the model on the right. Considering they are all taking place in small recesses, I’m going to lean hard on the AO as the source of the blame.
What I’m thinking is that the original high poly model could use some additional details, like bolts for the panels and maybe some additional 3D elements to raise the profile a bit more. Instead of insets, I’d like to add some outsets which might not suffer so much from the weird shadow artifacts that the AO is causing. What was missing from Mixer was “decals”. Whereas Painter had alpha masks that could be applied like stamps, Mixer didn’t have any good ones for sci-fi; if I wanted ancient Roman tablets, vines, moss, or graffiti, they’d have me covered, but the variation just isn’t there. Getting into the groove of creating my own is just more work that I’m not willing to take on at this point, considering how much I’m chewing on at the moment.
This is usually the point where I throw up my hands and whisper “fuck it” for another 364 days and shelve the hobby. I’ve done it so many times now it’s become part of the routine, along with going right back to it after conveniently ignoring the fact that I threw up my hands and whispered “fuck it” 364 days prior. I’ll try to make something of the high-poly version, but I might just hold off on the whole texturing thing until I can get my head completely around the baking process. I actually understand it now, at least for the measly four maps I need for my workflow (normal, AO, curvature, and material IDs), but as it seems to be the place where otherwise hidden issues come to light, I think I’d rather work on getting my modeling — and hopefully creativity — on track before I throw myself under the proverbial bus again.