blockydob.blend

Woof.

Late February. You are sick for the second time in the calendar year, which you think is bullshit. Whilst languishing melodramatically, you recall that, having spent much of the previous month on a 3D video game, you’ve been wanting to get back into the medium yourself. (The, uh, 3D part. You know fuckall about game-making.)

As you have nothing better to do than zone out to social media and YouTube videos, you stumble upon an entire low-poly character modeling course.

Your fate is sealed.

Your first venture is of course your own character. (You could follow the tutorial’s demo character, but you’ve done this before, and would rather wing it.) It is a good re-introduction to the program and how pixel-pushing in three dimensions works. You have no particular ambitions with it, but shiny new know-how that you absorbed from sicktime tutorial binging is already making your workflow much, much faster than the first time around.

Then you get lost in clothing confusion modifier hell. The file is set aside.

For your next attempt, you consider jumping off the deep end. Your actual character fixation du jour is some loser with blue hair and war crimes, and for convoluted reasons you will not be elaborating upon, you have taken to likening her to a dog. Specifically, this dog.

So, fuck it. Furry beam.

You get about this far before realizing

1. you have no idea how the face edge loops work,
2. you have no idea how the hand/paw edge loops work, and
3. you’ve toyed with this high-poly nonsense before, to a limited degree, and… well, you didn’t hate it or anything. But you also remember it taking forever, and that was only a head. You don’t really want to take forever on this.

You’re also done typing in second-person, now.

Get blockified, idiot.

The anatomy immediately & deliberately changed from “dog standing on hind legs”-type anthro to “human with dog head & extremities”-style structure, mostly because I thought the latter would be simpler to model. (The legs. I’m talking about the legs. That seemingly-backwards bend in the dog-type legs was a pain in the ass to position right, and I have yet to mess with moving limbs that way in 3D.)

And it was, in fact, simpler! By the end of the day I had the whole thing fully made and done, all cubular.

The markings weren’t particularly planned, but I thought a Minecraft-esque pixel style would fit the simplification nicely. So I did some quick automatic UV unwraps, scribbled on a texture to get a feel for how it might look, and called it a night.

Theoretically, the mouth was supposed to open. Some face and edge futzing later, that got done. And then I was extra-possessed and decided to make the entire model one (1) continuous object.

Would it have been easier to just make the legs their own objects & emphasize the disconnect to mesh with all the other split parts better? Probably. I like the connected look better, though.

Moving on!

90% of the base model has the exact same solid color. (Gradient textures might be neat to mess with some time, but not this time.) Did I need to line up almost every pixel on almost every side, ensuring all the while that the sides are distinct & the UV maps don’t overlap? No.

And yet!

There was a lot of back-and-forth nonsense between Blender and FireAlpaca (art program) here, btw. I thought it’d be clever to keep each part on its own layer, even if this meant re-exporting the PNG a billion times. At some point later in I gave up on this entirely, but an attempt was made.

Clothes got modeled along the way, to the extent that they’re modeled separately at all. As simplified from the original design: separate top & shoe cuffs, and everything else is part of the base texture.

These, I did keep snapped to the grid, all right-angled and front-back vertices overlapping. The shoulder area especially got simplified, compared to the confused tri-quad fuckery of the base model. I knew, from some other tutorials, that it should be possible to move both clothes and base model together seamlessly..ish. I probably could’ve gone back and simplified the base shoulders, too, but w/e I had to call it #GoodEnough and move on.

There was, at this point, a small dilemma. I hadn’t decided on exact base colors, and thought it might be fun if I could edit those directly in Blender. I also knew the not-separately-modeled clothing would be on an overlapping 2D-layer, and although I had (and still have) no real plans to swap it out, it’d be neat if that’s a thing that could be done. And, I wanted certain details like the metallic trim to use a different texture entirely, react a little differently to light. (The tutorial uses a flat, shadeless look, but I like more control over the light source and mood.)

In other words: want texture to Have Layer.

I did not figure this out all by myself, but sadly did not save links to whichever specific tutorials or stackexchange threads helped me out. RIP. In any case, this ↑ is what I landed on for a preliminary run…

…aaand we landed here! Some tweaks were made along the way: decided to just use one full-color image for the base texture, simplified the ear polygons, and the belt-metal got split off. An unintended side effect of how I set up the textures is that (not shown) the clothing, particularly on the ankle, shines very blue in the light. The tint is not supposed to be that strong. Oh well.

I also considered throwing in some subtle anti-aliasing to round out the markings and buttons, but quickly realized that if I wanted that to look consistent I’d need to anti-alias the rest of it, and I wanted to start making this sucker move.

It’d been a week, at this point, of doing basically nothing else with my spare waking hours. Perfectionism gnawed. It is good to remember, with any novel pursuit, that one’s first forays into the thing are always going to be the “worst.” It does not need to be Super Shiny Professional Polished 64. It only has to be done.

Finished, not perfect. Aaaaaaaaaaaa.

Not that I’ve been consistent with my file version breakpoints ever, but about 11 of the 32 .blend files constructing this thing were devoted entirely to rigging. Y’know, the model inside the model that actually animates. The rig. The process of rigging. Making the rig. And bugtesting the rig. And fixing the rig. And tweaking the rig. And—

An annoying thing with rigging is that it’s A Lot Of Tedium and not the fun kind. (Er, I think certain kinds of tedium are fun, anyway. Coding, for instance. Creative problem-solving. Mindless pixel-pushing, even. Rigging is, in my experience so far, mostly just clicking points in ways that should make them do the thing, and then trying to reverse-engineer why they are not doing the thing.) It’s also not super amenable to being shown visually.

Look, uhh, here’s the Crashsune tutorial segment on weight painting. That’s the bit I’m complaining about, mostly. I am extremely glad that this tutorial introduced me to Rigify, which auto-generates all the fancy bone movement bits and eliminates that entire chunk of tedium—even though it introduces its own overwhelm, such as, uh…

I do feel a bit like this is the equivalent of using Bootstrap and a CMS for a static two-page site—sorry, in plain less-nerdy English, it’s a bit like going “I would like to play piano,” being directed to a piano, and you can start plinking away but in the process of learning how to play well you also need to know exactly how each string is tuned, the wood the piano was crafted from, how the angle of the thingamabob affects the doohickey, and also if you want to play one or two specific notes you need to carve them out from scratch by ripping the material from a tree raw with your bare hands.

I did the equivalent of duct-taping sticks to a pre-built sandcastle and moved along, satisfied with my efforts and ready to make the silly little animations of my dreams oh wait no I didn’t

Remember how I drew attention to the eyes (lack thereof) in the planning phase? This is the part where it comes back to haunt me.

You see, the OG tutorial series uses flat textures for the entire face. This is simple and easy to accomplish, because the character’s face is mostly a single flat plane, anime-style. There’s a later addendum in which the irises are made moveable, which is also (relatively) simple and easy to modify, because the entirety of the eyes fit onto one (1) plane.

This is not the case for me.

I did, demonstrating some stray crumbs of forethought, make the eye area of the base texture transparent, and use a separate object underneath for the sclerae. Did I account for the multi-planar nature of the thing no. Did it take one simple easy try to get it right no. Did I manage to make the irises moveable, in a way that made sense from both front and side views, with nice smooth animation and everything oh. Surely you jest. (read: No.)

Eventually. Eventually. Several false starts and glitched-to-all-hell attempts later. I settled on making the entire* eye area (*inner! the brows are a different part!!) a flat, swappable texture. It basically follows the tutorial’s face-swap technique but just for the irises. I didn’t love the node redundancy, though, so after scavenging elsewhere and the tutorial’s comments, I found this stupid simple method from @gibbhartin643:

Now I have to redraw the texture every time I want the eyes to do something different, but. but. it does “work.”

Also I never applied the Mirror modifier so I’d have to re-wrangle everythign to make the eyes not exactly the same on both sides lol anyway—

Step one: Make boat.

Step two: Make the rest of the fuckign render.

There were a couple scrapped attempts at different lighting setups & some slight movement (since, y’know, it’s a fully-rigged model that can animate), but I truly was nearing the Sick Of Looking At This point and didn’t want to end on a bitter note. Ultimately I settled on lighting that attempts to mimic the 3D Viewport’s material view, threw in a slight side-shifting camera movement (which, hey, is the exact effect I talked about (but didn’t know how to implement) last time…!), aaaand…

Good fucking lord I learned a lot with this. Have some notes-to-self.

I am not not tempted to go back and redo some parts that are irking me here (face. but also smaller details around the joints, maybe add that antialiasing; the textures ended up a nightmare with multiple semi-overlapping images, and can absolutely be simplified) but. but but but. All of that would involve non-small & possibly-destructive changes to the model that I Don’t Want To Do Right Now. They are things that I am filing squarely under Learning Experience, will keep in mind moving forward, and whether I revisit this with improvements or (if enough time has passed & I’m substantially better at 3D by then) from scratch is up to future me.

Current-me out.

Overall time taken, not including preliminaries, concept sketches, or this page: a little over 40 hours. yeagh....

Have some screenshots that didn’t fit anywhere above.

Thanks for reading~