The art that draws itself
The most useful exercise I have done on the art side was also the least creative: walk the entire game, screen by screen, and write down every single thing that renders a graphic. Then mark each one free — meaning it draws itself in code — or paid, meaning somebody has to make a picture.
The result was clarifying in a way I did not expect. Almost everything your eye actually rests on is free. The paid list, when it finally emerged, was short, specific, and cost about the price of a sandwich.
The land is five sheets of glass
The depth you see while the wagon travels is not a painting and not a photograph. It is five separate planes, each drawn in code, sliding past at different speeds.
- The sky is a vertical gradient that crossfades between palettes as the run moves through its acts. Prairie light is not mountain light is not winter light, and none of them is a file on disk.
- A glow band sits over the sky at the horizon — warm, low, the light of a day that is either starting or ending and never quite committing to which.
- Three ridge lines — far, near, and foreground — are the same repeating shape drawn at three darknesses, scrolling at three speeds. That is the entire parallax. Three copies of one ridge is enough for a human eye to accept distance, which is a fact about human eyes rather than about ridges.
- A winter overlay is a white screen-blend that fades up as the run climbs, so the world drains of colour on a slope rather than switching at a boundary.
- Graves slide past in the near field as small drawn markers, each one belonging to somebody a run lost.
That last one is the only element on the list with anything to say. Everything else is scenery. The graves are the game quietly keeping score in the landscape, and they cost as close to nothing as makes no difference — a shape, a name, and a position that scrolls.
The whole shadowbox is the largest visual element in the game and its art budget is zero. That is not thrift for its own sake. It is that a gradient can crossfade and a painting can't. The land changes continuously across two thousand miles because it is made of numbers, and no amount of money buys that from a set of background plates.
Weather is a lens, not a picture
The temptation with weather is to draw it: a storm background, a snow background, a dust background, and a switch that swaps them.
Weather here is a tint over the one procedural land. Clear is a fair, neutral pass. Storm darkens everything and pulls the colour down. Snow ramps the white overlay. Drought warms the whole field toward dust.
Two consequences follow, and both are larger than they look. The first is that the weather art budget is zero and will remain zero however many weather states get added, because a new state is a new set of numbers rather than a new set of files. The second is that weather composes with everything else automatically. A storm in the mountains looks like a storm in the mountains, because it is the mountain palette with the storm lens over it — not a storm plate that has no idea where it is.
There is a lesson in that beyond art. When a system is a modifier rather than a variant, you get the combinatorial cases for free. When it is a variant, you get to make them all by hand, and you get to keep making them every time either axis grows.
Icons that are already there
Every meter in the game has an icon: grain for food, a flame for spirits, a wheel for the wagon, a heart for the body, and a drawn ox. Five small shapes, drawn inline as vector paths.
The interesting decision is what happens with goods. There are dozens of purchasable items across the shops — flour, salt pork, coats, whiskey, a fiddle, an axle, grease, a medicine chest — and the obvious approach is dozens of little item pictures.
Instead, every good routes to whichever of the five meter icons it touches, colour-keyed accordingly. Flour is the food icon. A coat is the health icon. A spare axle is the wagon icon. The goods-icon budget is zero, permanently, and adding an item to a shop requires no art at all.
This is a real trade and worth being straight about. You lose the small pleasure of a shelf full of distinct objects; a fiddle deserves to look like a fiddle. What you gain is that the icon tells you what the item does — a player looking at an unfamiliar good sees immediately which meter it feeds, which is the only thing they need at the moment of buying. Legibility beat charm. On a phone, legibility usually does.
Where the money actually goes
After the audit, the paid list was genuinely short: illustrated scenes for the encounter cards, and portraits.
The scenes are the easy half. Cards that depict the same kind of moment can plausibly share one illustration — the four different ways a wagon breaks are all, visually, somebody kneeling in the dirt beside a wheel; the several ways a person falls ill are all somebody laid up under canvas. Folding those brought forty-five cards down to about twenty-nine images, and scenes are forgiving: no two of them need to be consistent with each other, so a little variation in style reads as variety rather than error.
The portraits are the hard half, and they are the only real risk in the whole plan. Nineteen faces — the ten people you can bring and the nine shopkeepers you meet — and a human eye tracks faces with a precision it applies to nothing else. Twenty ridge lines can vary and nobody notices. Two portraits lit differently and the whole cast looks assembled from three different games.
So the plan for portraits is the same framing and the same lighting description on every one, and — more importantly — three test images judged hard before committing to the batch. That is the actual discipline: not the consistency instruction, which is easy to write, but the willingness to look at three and stop.
The total for everything, at reasonable quality, comes to a little over three dollars. That number is not a boast about cheapness. It is what the audit was for: knowing exactly which fifty-two images matter means you are not paying for the four hundred that don't.
The upgrade layer
Here is the thing I would tell anyone building a small game with no art budget: the procedural version is not the placeholder. It ships. It is the game.
Everything on the paid list is an upgrade layer — it swaps small vector emblems for illustrated scenes and silhouette heads for real faces. You can ship none of it, some of it, or all of it, and at every point in between you have a complete game that looks deliberate rather than unfinished.
That property is worth more than the art. A project where the art is a prerequisite has a cliff in the middle of it, and a lot of small games die on that cliff — not because the art was unaffordable but because the game could not be seen, played, or judged until it existed. A project where the art is an upgrade has no cliff. It has a working game from the first week, and a slider you can move later when you know which parts deserve it.
The gradient, the three ridges and the tint are not what I would have chosen with unlimited resources. They are what let the game exist early enough to find out whether it was worth illustrating.