Segmentation takes a single mesh and divides it into separate parts — a chair into seat, back and legs; a character into head, torso and limbs. Each part comes back as its own mesh.
Why split a model at all
- Separate materials. Different parts take different textures without a shared UV layout fighting you.
- Animation. A part is a thing you can move independently — a lid, a wheel, a drawer.
- Selective editing. Replacing one component without touching the rest.
- LOD strategy. Simplifying the parts nobody looks at while keeping the ones they do.
Where the cuts land
Segmentation reads shape, not intent. It cuts where the geometry suggests a boundary — a change in curvature, a narrow join, a distinct volume. That usually agrees with how a person would divide the object, and sometimes does not.
It has no idea what the object is for. A handle that is visually continuous with a body stays attached, even though functionally it is a separate part. A decorative ridge that reads as a distinct volume may come back as its own piece, even though nobody would call it a component.
Segmented is not the same as modelled in parts
The pieces meet exactly where the surface was cut. There is no hidden geometry behind a join, because there was never anything behind it — a segmented drawer has no back face where it met the cabinet, and opening it reveals a hole.
If parts need to move relative to each other and show their interiors, they need to be modelled that way, not segmented that way.
Working with the result
Parts are usually easier to handle as one file with the pieces kept separate underneath a single parent, rather than as a folder of individual meshes. That imports as one object that is already divided, which is what most engines and DCC tools want.