Across all the systems in this series there’s one rule that decided most of the arguments: when there’s a choice to make, pick whichever option looks better. Not the most realistic option, not the most technically interesting one — the prettier one. That sounds obvious, but taking it literally and consistently changes a lot of small decisions, and the small decisions are what add up.
Beauty as the spec, not the polish
It’s normal to build something functional first and make it look nice later. This project inverted that: how good it looks was the actual goal, and correctness was in service of it. A river system isn’t there because simulating hydrology is neat; it’s there because rivers and waterfalls make a landscape worth looking at. The question for any feature was always “does this make the world more pleasant to be in,” and if the answer was no, it didn’t matter how clever it was.
What the rule actually changes
Stated as a principle it’s vague, so here’s how it showed up in practice:
- Bias the random ranges toward good-looking results. Procedural generation rolls a lot of dice. Wherever it does, the ranges are tuned so the likely outcomes look good, rather than allowing the full range and accepting that some worlds come out ugly.
- Make the striking conditions common enough to actually see. Sunsets, night, fog, fireflies, glowing water — if these only happened rarely you’d never experience them. They’re tuned to occur often. A feature you don’t see might as well not exist.
- Curate where raw randomness looks cheap. The house colours come from hand-picked palettes, not random RGB. The notes come from a scale that can’t clash. Generate the structure freely, but constrain the parts where random choices tend to look bad.
- Prefer smooth to sharp. Soft transitions between biomes, blended foam lines, gusts that roll across a field instead of snapping. Hard seams read as artificial; gradients read as natural.
- Spend a little more when it clearly pays off. If something can be made noticeably nicer for a modest extra cost, do it. Volumetric fog for god-rays, reflections on the water, a curved grass blade up close — none essential, all worth it.
Why a rule helps
The value of having one stated principle is that it settles trade-offs quickly and consistently. Faced with two ways to do something, you don’t relitigate your priorities each time — you ask which looks better and move on. Over hundreds of those small decisions, a consistent bias toward the prettier option is most of what separates a tech demo from somewhere you actually want to spend time.
It also kept the project honest about what it was. This was never meant to be a realistic terrain simulator or an efficient engine showcase. It was meant to be a nice place to walk around. Keeping that front and centre is the reason all the systems in this series point in the same direction.
The last post is about the payoff: stepping inside one of these worlds in VR.