THE POLYLUSION ART STYLE: WHY WE CHOSE LOW POLY

I spent about three weeks back in 2023 trying to model a single tree.

It was supposed to be a hero asset for a prototype I was working on, some little forest exploration thing that never saw the light of day. I had Blender open, I had a stack of reference photos pinned to my second monitor, and I had a YouTube tutorial paused on a frame showing a guy adding the seventh layer of bark detail to his oak. The tree had something like 80,000 triangles by the time I gave up. The bark looked great in close-up. The leaves looked great if you squinted. From any reasonable game distance, it looked like a green smudge sitting on top of a brown smudge. Three weeks. One smudge.

That tree is the reason Polylusion looks the way it does.

I want to talk about why every game we put out leans into flat-shaded geometry, hard edges, and color palettes that look like they were ripped out of a 1980s board game box. Some of this is practical. Some of it is taste. Some of it is, honestly, a survival strategy. But all of it adds up to a deliberate choice, and I think it's worth writing down somewhere instead of just answering it in Discord DMs every time someone asks.

The solo dev math

Here's the part nobody likes to admit out loud. I am one person. I have one brain, one set of hands, and a finite number of hours in a week before my wife starts giving me that look. When you are a solo developer, every art decision you make is also a time decision, and every time decision is a release-window decision. Photoreal art is not just expensive in dollars. It's expensive in years.

A photoreal tree, modeled and textured to a standard that wouldn't embarrass me on a Steam page in 2026, is probably a two-week asset if I'm being honest about my skill level. A low poly tree, the kind I make for our games, is forty minutes. Sometimes less if I'm cribbing from a kitbash I already built. That's not a small difference. That's the difference between shipping a game with a forest in it and shipping a game with three trees and a lot of fog.

Multiply that across every prop, every character, every environment piece, and the gap becomes absurd. A photoreal version of one of our games would take me, conservatively, eight to ten years to make alone. I don't have eight to ten years per game. Nobody who eats food does.

The low poly style isn't a compromise I made because I couldn't do photoreal. It's a system I chose because it lets me actually finish things. Iteration speed matters more than fidelity when you're shipping. I'd rather playtest a rough idea three times this month than polish a beautiful idea I never get to test.

File sizes and the mobile question

There's a boring technical reason too. Our builds are small. Like, really small compared to most modern games. A low poly model with a flat-shaded material and a 256x256 texture (sometimes 64x64, sometimes literally a gradient on the vertex colors) is maybe 30 kilobytes. The photoreal equivalent, with PBR maps for albedo, normal, roughness, metallic, ambient occlusion, and whatever else the engine wants this week, is closer to 30 megabytes. That's a thousand times bigger for one prop.

When I want to ship a game on mobile, or web, or just have a download that doesn't take twenty minutes on a hotel wifi, that ratio matters a lot. Our entire current project fits in less space than the launcher for some AAA games. That's not bragging. That's just what happens when you stop fighting the medium.

It also means the games run on garbage hardware. My nephew plays our stuff on a hand-me-down Chromebook that has no business running anything, and it runs at sixty frames per second because there's nothing for the GPU to choke on. I like knowing that the kid with the cheap laptop gets the same experience as the kid with the RTX card. That's a weird hill for me to die on, but here we are.

The aesthetic argument, which is the real one

Okay, the practical stuff is real and I genuinely care about it. But if I'm being honest with myself, I'd probably still make low poly games even if I had a team of forty and a budget that let me ignore file sizes. Because I think it looks better. Or, more carefully, I think it looks more like itself.

Photoreal art ages weirdly. You can pick up a game from 2014 that was the best-looking thing on the planet at launch, and now it looks like a wax museum. The lighting is wrong by current standards. The hair tech is wrong. The faces sit somewhere in the lower part of the uncanny valley where everyone looks vaguely concussed. Photoreal is always reaching for a target that keeps moving, and the moment the target moves, the work looks dated.

Low poly doesn't have that problem because it isn't pretending to be anything except itself. A flat-shaded triangle is a flat-shaded triangle. It doesn't try to fool you into thinking it's a leaf or a person, so it never fails at being convincing. It's stylized in the same way a comic book or a children's picture book is stylized, and those don't really age. A Dr. Seuss illustration from 1957 looks exactly as good now as it did then. That's the club I want to be in.

There's also the uncanny valley thing, which I take seriously. Faces especially. I've watched my own face get rendered in real time in some of these new engines and it's deeply upsetting. Something about the eyes. With low poly, a face is a few flat planes, maybe two dots for eyes, and your brain fills in the rest. It's the same trick a stick figure pulls. You aren't asked to believe it's a person. You're asked to accept it represents a person, and your imagination does the heavy lifting from there. That's a much friendlier deal for the player.

When you take detail off the table, what's left is shape and color. And shape and color are where the actual emotional content of an image lives anyway. A great low poly scene reads as a great composition first, with no surface noise to hide behind. If your silhouette doesn't work, you can't paper over it with a 4K normal map. That forces me to be a better designer, which is good for me, and it forces every frame of the game to be legible from across the room, which is good for the player.

The strategic angle

Now for the cynical part, which I'll own. Low poly is having a moment, and we're riding that moment on purpose.

If you scroll through Steam right now, the homepage is a slurry of brown and gray photoreal screenshots that all blend together. Even the good ones are hard to tell apart from a thumbnail. A low poly game pops out of that grid. It looks different. Different is the most valuable thing you can be in a marketplace where the average user makes a buy-or-skip decision on a tile that's about the size of a postage stamp.

Crossy Road figured this out years ago. The game is a Frogger clone, mechanically, and there are a thousand Frogger clones. But Crossy Road has those chunky little characters with their voxel jiggle and their pastel cars, and you recognize a screenshot of it from a hundred yards away. That recognizability is the whole product. You could rebuild the same game with photoreal frogs and it would not have made a tenth of the money, because nobody would have remembered seeing it.

A Short Hike is the one I keep coming back to. Adam Robinson-Yu made an entire melancholy little open-world game with low poly characters and chunky pixel filtering on top, and it's one of the most emotionally resonant indie games of the last decade. The art style isn't a limitation. It's the entire personality of the thing. Take it away and you have a worse game.

Tunic too, even though it's slightly more polished than what I'd call true low poly. The isometric framing, the chunky shapes, the way the little fox catches the light. Andrew Shouldice spent seven years on that game and it looks like nothing else on the platform. That's not an accident.

And then there's everything Lucas Pope has ever made. Papers Please, Return of the Obra Dinn, the new one. He's not exactly low poly in the same sense, but he's working in the same tradition: pick a hard aesthetic constraint, commit to it completely, and let the constraint become the identity. Obra Dinn is a 1-bit dithered murder mystery. There is no universe in which a photoreal version of that game is more interesting. The constraint is the magic.

I want our games to belong on that shelf. Not the same as those games, but in the same conversation. The constraint is the magic.

What I learned from the bad tree

Going back to that tree from 2023 for a second. The reason I gave up wasn't that I couldn't make it look better. I probably could have, given another month. The reason I gave up is that I realized I was solving the wrong problem. Nobody was going to look at that tree and feel anything except mild appreciation that the bark was detailed. Detail is not the same as feeling. Detail is just detail.

A low poly tree, by contrast, can carry feeling because every choice you make about it is loud. The angle of the trunk, the color of the leaves, the way the silhouette breaks against the sky. There's no detail to distract from those choices. Everything that's there is there on purpose, and the player's eye catches every decision. That's much more interesting to me as a maker than chasing the bark texture demon.

If you want the longer version of how I think about this, I wrote a whole thing about the best low poly games and why they work that gets into more specifics. Some of those games are direct influences on what we're doing. Some of them are just games I love. The line between those two categories is pretty blurry for me.

The workflow that fell out of all this

What I ended up with, after a couple of years of refining, is a pretty tight pipeline. I model in Blender with the snap-to-grid turned on, hard edges marked everywhere, no smooth shading anywhere. Most assets get vertex colors instead of textures, which means no UV unwrapping, which is the part of 3D art that makes me want to throw my keyboard. When I do use textures, they're tiny and palettized, often hand-painted in like ten minutes.

Animation is mostly rigid. Characters have a head, a body, and maybe four limbs that bend at one joint each. No fingers. No facial bones beyond a mouth that opens and closes. It looks great because the constraints are consistent. The moment you put fingers on a low poly character, you've broken the visual contract and now everything else looks weak by comparison. Pick a fidelity level and stick to it across every asset. That's the rule.

Lighting is similarly minimal. One directional light, one fill light, a flat ambient, and we call it a day. No real-time shadows on most things because they're expensive and they fight with the flat shading. The whole scene reads like a diorama, which is exactly the vibe I'm after. A small handmade world that you can pick up and turn over in your hands.

This whole pipeline, end to end, lets me ship a new prop in under an hour, a new character in maybe a day, and a new level in a week. Compare that to the photoreal version of any of those numbers and you understand why we are where we are.

Why I'm writing this down

I get asked about the art style more than I get asked about anything else. People want to know if I'm going to "upgrade" eventually, like low poly is a phase I'll grow out of. And I want it to be on the record somewhere that no, this isn't a phase. This is the whole plan. Every game we make is going to look like this, give or take, because I've thought about it a lot and I keep arriving at the same answer.

The tree from 2023 is sitting in a folder on my drive labeled "do not delete." I don't open it much. But every once in a while when I'm tempted to add one more loop cut or one more texture map to something, I pull it up. Eighty thousand triangles. Three weeks. One smudge.

I'd rather make a hundred trees that look like trees than one tree that looks like a tragedy. That's the whole pitch. That's why everything we make has hard edges and bright colors and faces with two dots for eyes. It's not a limitation. It's the point.