An algorithm can create a place or pattern on demand rather than having someone create each part by hand. Its parameters can control broad features such as the roughness of terrain.
Results can be checked for consistency at boundaries and for how well they fit a desired style. Developers can combine rule-based methods with learned components when both contribute useful properties.
ELI5
For example, a game can use a set of rules to generate different patterns for stone walls.
A designer can change the rules to make the patterns coarser or more regular.
What makes a generated region repeatable?
Fixed parameters and a fixed random seed can produce the same result when the generation process is deterministic.
Can it use trained models?
Yes, learned components can work alongside explicit rules, provided their outputs fit the wider generation process.


