Quads vs triangles: when topology matters and when it does not
Every GPU renders triangles, so why do artists insist on quads? A plain explanation of topology and when it is worth fixing.
By the threeD editorial team. Published . 7 min read.
Topology is the way a mesh’s vertices, edges and faces are connected. Two meshes can have exactly the same shape and very different topology. Most of the time a viewer cannot tell the difference. It starts to matter as soon as you try to edit, subdivide or animate the model.
Everything becomes triangles in the end
Graphics hardware draws triangles. When you export a quad-based model to a game engine, it gets split into triangles, either by your exporter or by the engine on import. So for a static model that will only be rendered, triangles are completely fine, and decimated triangle meshes are the norm for props, scans and background assets.
The case for quads is about what happens before rendering.
Why artists prefer quads
- Subdivision. Subdivision surfaces, used for smooth high-poly models, give predictable results on quads. Triangles and five-sided poles produce pinches and bumps.
- Edge loops. On a quad mesh you can select a whole ring of edges with one click, then add, slide or remove it. This is how artists add detail where it is needed and remove it where it is not.
- Deformation. When a mesh bends at a joint, evenly spaced quads arranged in loops stretch and compress smoothly. Irregular triangles crease and collapse.
- UV unwrapping. Clean loops give natural places to put seams and make unwrapping easier.
Edge flow
Edge flow is the direction the loops run across the surface. Good edge flow follows the form and the way the model will move: loops encircle the eyes and mouth, run around the shoulder and wrap around elbows and knees. When the arm bends, the loops at the elbow fold neatly instead of tearing the surface.
AI-generated meshes and 3D scans have no edge flow at all. Their triangles are placed to describe the shape, not to support animation. That is fine for a statue and a problem for a character.
Does your mesh need retopology?
| Situation | Recommendation |
|---|---|
| Static prop in a game or AR viewer | Decimate to a triangle budget; no retopology needed |
| 3D print | Keep it dense and watertight; topology does not matter |
| Character that will be rigged | Quad retopology, at least around joints |
| Model you will keep sculpting or subdividing | Quad retopology |
| Hard-surface model you will edit (bevels, booleans) | Quads, or remodel by hand for precision |
In threeD’s remesh tool this is the choice between triangle decimation and quad retopology. The live demo on that page shows decimation running on a real mesh.
What automatic retopology can and cannot do
Automatic quad remeshing follows curvature well and produces even, quad-dominant meshes that subdivide cleanly. It is less reliable at the artistic decisions: exactly where the loops around a mouth go, or how many loops a shoulder needs for an extreme pose. For hero characters, treat automatic retopology as a strong first pass and check the face and shoulders by hand.
Terms worth knowing
- N-gon: a face with more than four sides. Fine on flat, static areas; avoid on anything that bends or subdivides.
- Pole: a vertex where three, five or more edges meet. Unavoidable, but best placed in flat areas that do not deform.
- Manifold: a mesh where every edge belongs to exactly two faces. Needed for 3D printing and many modelling operations.
- Decimation: reducing the number of faces while preserving shape, without regard for edge flow.