IT lexicon › Programming › Level of detail (LOD)

Level of detail (LOD)

Programming På svenska → Updated: 2026-10-24

Swapping an object's geometry for a simpler version the farther it sits from the camera, so the renderer never spends effort on detail nobody can see.

James Clark formalised the idea back in 1976: keep several versions of each model — full resolution up close, coarser meshes at range — and pick one by distance or on-screen size. The payoff is large, because distant objects cover few pixels yet cost just as much to draw in full. The final rung is often an impostor or a flat sprite.

Discrete LOD snaps between prebuilt meshes and shows a visible "pop" at each switch; continuous LOD morphs smoothly but costs more. The technique pairs with frustum culling and occlusion culling — culling throws away what you can't see, LOD simplifies what you can. Unreal Engine 5's Nanite (2021) virtualises geometry and chooses detail per triangle cluster, largely retiring hand-authored LOD rungs.

← Back to the lexicon