Basically, the idea is that we want, for example, four levels of detail (LOD), with objects closest to the camera having the highest detail. But it takes time to generate models from their abstract description. So the solution is to generate the model at the lowest level of detail, and then fire off a thread to generate the remaining three levels of detail. The resulting image looks worse, but the animation is more fluid. It’s preferable to a three-second pause in the middle of a combat scene. Visually what you see is a 3D model that increases in quality with time, much like a progressive JPEG.