KAIJU PROTOCOL DEV BLOG: DESIGNING THE MONSTERS

I thought designing kaiju would be the easy part. Big monster, cool teeth, done. Put scales on it. Give it a roar. Done again. Then I sat down with a sketchbook, tried to draw the first monster for Kaiju Protocol, and realized I had absolutely no idea what I was doing. Four months later I have a pipeline that works, three finalized monsters in the game, and a much deeper respect for the people who designed Godzilla in 1954 with rubber suits and zero iteration tools.

This is a dev blog. No marketing spin, no trailer beats. Just notes from a solo dev on why designing monsters is harder than it looks, what I got wrong the first time, and what I actually do now to take a kaiju from a scribble on paper to a thing that shows up on screen and feels like a character.

Why monsters are harder than they look

The trap is that monsters seem like free design space. You can do anything. Six legs, four eyes, wings on the back, beam from the chest. Nobody is going to argue the biology. You are not shackled to a reference photo of a real horse. The creative constraints look loose.

That is exactly the problem. When everything is possible, nothing reads. A monster with six legs, four eyes, wings, horns, a tail, armor plates, and a chest beam is not scary or cool. It is noisy. The eye does not know where to look. The silhouette is a blob. You cannot describe it to your friend after the fight.

Great kaiju are the opposite of that. Godzilla is a stack of plates on a reptile body. King Kong is a gorilla, bigger. Mothra is a moth. The first time you see them you already know what they are. You can draw them from memory after a single viewing. That is the bar, and I did not understand how high that bar actually is until I tried to clear it.

My first monster was a crab-spider thing with a crystalline carapace, hooked legs, a whip tail, and eyes in places eyes should not be. On paper it looked fine. In engine it looked like a pile of laundry. At fifty meters the player could not tell it apart from the environment rubble. At five hundred meters it was a speck with too many little bits. That failure is what sent me back to first principles.

Silhouette first, always

Rule one, which I learned the hard way. The monster has to read as a silhouette. Not just from the front. From every angle. At every distance. In every lighting condition the game throws at it.

The test I run now is simple. I take the monster's 3D blockout, drop a pure black material on it, and put it in front of a white background. Then I rotate it through 360 degrees on a turntable and take a screenshot every 15 degrees. Twenty-four silhouettes. If I cannot identify which monster it is from any of those silhouettes, the design is not done.

I also do a distance test. Same black on white setup, but I render the silhouette at four distances. Close, medium, far, and almost-a-dot. At close range the player needs to read the details. At medium range they need to read the pose and the main mass. At far range they need to read a unique shape. At almost-a-dot they need to read a unique outline, something as simple as does this thing have wings or not.

Godzilla passes all four of those tests. The stacked dorsal plates are visible as bumps on the back even when Godzilla is a silhouette the size of your thumbnail. The long heavy tail, the upright posture, the short arms, those are all still legible when every detail is gone. Kong passes too. So does Mothra. That is not an accident. Those designs have been refined across dozens of films and they have converged on shapes that work at any distance.

My crab-spider monster failed this test on every axis. From the front it looked like a spiky blob. From the side it looked like a different spiky blob. From far away it was a dot with some speckle around it. I was designing by adding detail, which is the opposite of designing by silhouette. Once I forced myself to start from a filled black shape and only add details that survived the distance test, everything got easier.

In practice this means I start every monster with a silhouette pass. No anatomy. No textures. Just a painted black shape on a canvas. I push that shape around until it reads as a thing you have never seen before but you could describe to someone with three words. Once I have that, and only then, do I start filling in what is inside the silhouette.

Every monster needs a verb

Rule two, which I stole from the Monsterverse and every fighting game I have ever loved. Each kaiju needs a signature verb. A thing only they do. A mechanic you cannot confuse with any other monster in the roster.

Godzilla's verb is the beam. Everything else follows from that. Heavy slow walk, close range swipes, tail spin, and then the beam as the signature long range tool. The whole silhouette points at the beam, because the spines glow before it fires. The design and the verb are the same design.

King Kong's verb is climb. He scales buildings, rips them apart with his hands, swings, brachiates. You do not see Kong shoot a beam because that would break the verb. The arms are long and powerful because the verb demands long powerful arms. The feet are hand-like because climbing demands grip.

Mothra's verb is fly. Not just flight as traversal. Flight as the whole kit. Mothra hovers, weaves, drops powder, generates wind. The wings are the signature silhouette piece and they are also the primary mechanic.

When I mapped this onto my own roster I realized my first three monster concepts all had the same verb, which was "hit things with claws at short range." Three melee brawlers. I had designed the same monster three times with different textures. Of course they felt interchangeable. They were interchangeable.

The current roster is built around verbs first. Monster one is a slow tank with a ranged beam, so the whole body is a tripod stance with a glowing chest cavity. Monster two is a fast fly-in glass cannon, so the design is lean, aerodynamic, with huge wings and almost no armor. Monster three is a burrower that tunnels through streets and erupts from below, so the silhouette is a fat segmented worm with no legs and a drill-like head. Three monsters, three verbs, zero overlap. That is the bar I am holding every new design to.

If I cannot describe a new monster's verb in four words or fewer, the monster is not ready. "Throws buildings as weapons." That is a verb. "Grapples with a long tongue." That is a verb. "Is really strong and also fast and has poison" is not a verb. It is a shopping list. Shopping lists do not make good monsters.

Color and texture for mood, not detail

Rule three, and this one took me way longer than it should have. In a low poly stylized game, color and texture are not there to sell realism. They are there to sell mood. The second the player sees the monster, the palette should tell them how the fight is going to feel, before a single move is made.

This is where I had to unlearn a lot of instincts from looking at hyper-detailed creature art online. Those monsters are painted with dozens of color variations, skin tones, subdermal veins, rim lighting, wet spots. In a 4K render, those details sing. In a low poly game running at 60 frames per second with a stylized look, those details turn into mud. The palette collapses. You end up with a brownish shape that is hard to read.

So I pick a tiny palette for each monster. Three main colors, maybe four if I really need a pop accent. And I pick those colors to communicate emotion, not biology.

Monster one, the tank with the beam, is deep charcoal grey with a hot orange glowing cavity. Two colors. The grey is inert, industrial, unkillable. The orange says danger, says hot, says something is loading. The player sees that orange glow start building across the body and they know they have two seconds to move. That is the palette doing gameplay work. Brown veins and skin texture detail would not add to that reading. They would subtract from it.

Monster two, the flying glass cannon, is pale blue with silver highlights and white wing membranes. Cold, fragile, ethereal. The palette tells the player this thing will die if they land a hit, but they have to land a hit first. Compare that to the charcoal tank. Without a word of UI, the player reads those two monsters as opposite playstyles.

Monster three, the burrower, is sickly yellow-green with bone-white spines. It is nature gone wrong. Infection colors. The palette is uncomfortable to look at, which is the point. Boss fights should have a feeling and the feeling for this one is "get away from me."

Texture follows the same rule. I am not trying to simulate skin. I am trying to sell a surface. The tank has faceted armor plates with hard edges and no detail in between, because rigidity is the mood. The flyer has smooth curves with a pearlescent shimmer in the wing texture, because fragility is the mood. The burrower has a wet-looking gradient with irregular dark blotches, because wrongness is the mood. If a texture detail is not serving the mood, it gets cut. Low poly discipline forces you to decide what matters.

The iteration process

Rule four is not a rule, it is a workflow. This is what actually happens, step by step, when I design a new monster.

It starts on paper. Actual paper, with a pencil, because I draw worse and think better on paper than I do in Photoshop. I fill a page with ten to twenty quick silhouettes, each one no more than a minute. Most of them are bad. That is fine. The point is to find two or three that have a shape I have not seen before. I circle those.

From the circled silhouettes I pick one and refine it. Still in silhouette only. I do four or five iterations of the same shape until the big mass feels right. Then I do a quick anatomy pass on top of the silhouette with a pencil, sketching what is inside. Legs, head, tail, whatever. This is the first time I am allowed to think about details.

Once the paper sketch feels locked, I move to Blender for a blockout. No textures. No details. Just primitive shapes pushed together to approximate the silhouette in 3D. A blockout should take me one afternoon. If it takes longer, the design is too complicated and I need to simplify before I go further.

Blockout goes into Unity immediately. Not after it looks finished. Immediately. I drop it into a test scene with a reference character for scale, a reference building, and our standard lighting setup. Nine times out of ten the monster looks wrong in engine even though it looked fine in Blender. Scale is wrong, proportions read differently in perspective, the silhouette breaks down at game camera distances. This is where most designs die or get major revisions. Better to find out now than after I have spent two weeks on textures.

Revisions go back to the blockout. Sometimes back to the paper sketch. Occasionally all the way back to the starting silhouettes and I pick a different circled shape. I have killed three monsters at this stage that I really liked on paper. That is the cost of holding the bar.

When the blockout passes the in-engine test, I do a retopology pass to build the final low poly mesh. Then texturing, which is a second chance to cut details that do not serve the mood. Then rigging, animation, and finally the verb gets implemented as a gameplay mechanic. By the time the monster fights the player, it has been through four or five rounds of revisions at different fidelities, and most of what I drew in the first week is gone.

This whole loop takes about three weeks per monster for me, solo, working evenings and weekends. I know people on full teams who do it in a week. That gap is real and I am not going to pretend otherwise. But three weeks is survivable, and the result is a monster that works, not a monster that looks cool in a portfolio shot and falls apart in play.

What I still get wrong

Plenty, honestly. My monsters are slower to design than I want. I still catch myself adding details before the silhouette is locked, and I have to force myself to delete them. The verb test saves me from boring designs but it also means I cannot use a lot of cool shapes that do not have a corresponding mechanic. Sometimes I mourn those.

Scaling is also harder than I expected. A monster that reads great at its intended size sometimes breaks when a boss fight calls for it to grow. A monster that looks menacing at fifty meters tall can look comical at two hundred meters tall, and the fix is rarely "make it bigger." If you want to see how much size changes perception of a monster, there is a whole other rabbit hole in our kaiju size comparison post that goes into it.

And I have opinions about the broader look of the game that bleed into every monster design. Our polylusion art style writeup covers that in detail. If you want the actual modeling workflow that takes a blockout to a final asset, the low poly modeling guide has the Blender side of it. And if you are curious how these monsters interact with the world they wreck, the destructible cities dev blog is the companion piece to this one.

Designing monsters is harder than it looks because every decision has to serve three masters at once. The silhouette has to read. The verb has to be unique. The mood has to land. Miss any one of those and the monster is forgettable no matter how much work you put into the scales. Nail all three and the monster becomes the kind of thing players draw from memory, which is the only real test.

I have two more monsters to design before the roster is locked, and I am already dreading them a little. But a lot less than I was dreading the first one, which is progress.