Vertices, edges and faces
Every polygon mesh is built from three elements. A vertex is a point in 3D space. An edge is a straight line joining two vertices. A face is the flat surface enclosed by a closed loop of edges, and that face is the polygon. Connected faces make the mesh, and the mesh is the model.
The number of faces is the polygon count, and how closely they are packed is the mesh density. Detailed areas such as fittings, handles and tight curves carry more faces; large plain surfaces need very few.
Tris, quads and n-gons
Quads, four-sided faces, are the working standard for modelling. Their edges run in continuous loops, so subdivision smooths them predictably and they deform cleanly when a model is animated.
Triangles are what real-time engines and formats such as glTF ultimately draw, so every mesh is triangulated at export or render time. N-gons, faces with five or more sides, are convenient on flat areas but can shade badly on curved surfaces and triangulate unpredictably, so artists keep them off anything that bends.
- Quads: the default for modelling, subdivision and deformation.
- Tris: what the renderer and real-time formats draw; fine where a mesh is final.
- N-gons: acceptable on flat, rigid areas; avoided on curves and anything that deforms.
High-poly and low-poly models
More polygons allow finer detail and smoother curves, at a higher computational cost. High-poly models suit photoreal stills and film, where every curve may be seen up close and render time is spent offline.
Low-poly models suit real-time uses such as configurators, 360 viewers, AR and games, where the scene has to redraw fast enough to stay interactive on the device. Surface detail from a high-poly version can be baked into a normal map, so a light mesh still reads as detailed.
Retopology: rebuilding a mesh for its job
Retopology means building a new, cleaner mesh over an existing one. It is used to turn a dense sculpt or scan into a model with sensible edge flow, or to make a lighter version of a product for real-time use.
Product work often needs both: a detailed model for stills and animation, and a retopologised version for a configurator or AR. Planning for both at the brief stage avoids building the product twice.
How many polygons does a model need?
There is no universal number. A hero still rendered for print can carry millions of polygons, while a model for a phone AR viewer has to stay small enough to download and draw smoothly. Decide the destination first, then put polygons where the silhouette and close-up detail need them.
When a brief says what the model is for, a studio can build to that budget from the start. When it does not, the usual result is a model built for the wrong use, then rebuilt.