The level of detail algorithm decreases the complexity of a 3D object representation as it moves away from the viewer or according other metrics such as object importance, eye-space speed or position.

When certain conditions are met, the mesh is split. The exact numbers can be changed in the preferences:

• The base mesh of the current object (base subdivision level) is not too dense.
• The current object is subdivided a couple of times.

When split, each piece will consist of those faces which are derived from the same base level face. So for example if the current object is subdivided to the fourth level, each piece will consist of 64 quads. This layout has the following advantages over the general layout:

• Each piece will have the same topology as the others, so there is no need to create a different index buffer for them, the same index buffer can be used for all the pieces. This saves video memory.
• Since the topology of the pieces is always a plain grid, Mudbox can reduce the number of rendered faces easily based on their distance from the camera. Rendering dense objects is therefore efficient, and allows for fully interactive camera manipulation, regardless of the mesh size. This also substantially reduces video memory usage.

In several cases the base mesh has too many faces preventing Mudbox to use this kind of layout. In this case a Mesh Division plug-in preprocesses the mesh, and tries to group faces on the mesh together and form a new quad from four small quads. These quads then can be further grouped together to form a bigger quad. If the topology of the mesh is not noisy, this method can reduce the base face count by about 90%, making it possible to use LOD with meshes with more than 10k faces.