One of the big barriers to entry into the procgen world is the complexity and steep learning curve.
How have you dealt with that in the past?
In Starlink: Battle for Atlas we built 7 (and then some) circumnavigable planets. And even though we kept the surface area of the planets super small (160km2) it was still a lot of ground to cover (pun intended).
Building the planets was a combination of multiple different techniques, a layered approach to asset placement, strict guidelines, etc.
It took a long time to get everyone into the same procedural creation mindset, and ultimately we came up with a compromise that worked for the project.
For an Artist, it was faster and easier to hand-build things from a concept, and recreating that concept with pure logic was an extremely daunting task.
We couldn’t simply place down copies, of the same template (something akin to a blueprint or prefab) as the repetition would be super obvious.
In comes in the Diorama, a procedural variation building technique.
1 . An Artist would build an original Diorama, from the concept. A diorama was a combination of assets in an aesthetically pleasing, artist-built asset.
For example: A Big Tree, small Trees, rocks, and some shrubbery.
The following is an example of a Diorama, well all of the small clumps are small dioramas, combined
Andrey Chernykh (Level Artist at UbiToronto)
Let’s say that the big tree is the “centerpiece” of the diorama. Instead of working out the rules of how shrubs and rocks were placed around trees and how the grass grows around all of it, an artist would actually manually create the original diorama, by hand.
That template then ran through logic that would modify the diorama, scaling, rotating, replacing with variants, deleting assets, translating. and therefore creating variations of that same diorama. Seeded by their world position.
Dioramas could be nested into larger ones, which would become biome pieces.
The placement of the dioramas would follow rulesets based on the Terrain, slope, splatmaps, etc.
Each asset would be snapped to the ground, with a parameterized alignment to the terrain normal.
This was a very fun project with huge ambition and a relatively small team that was able to pump out a lot of great quality content.