1
\$\begingroup\$

Currently I am a bit puzzled. I am writing my master thesis about procedural mesh and texture generation and for that I am writing a Blender Add-On to generate low poly planets with physical physically based rendering textures.

This blog (http://blog.digitaltutors.com/modeling-with-quads-or-triangles/) says that it's better to use quads instead of triangles and n-gons. The wireframe looks cleaner, subdivides are more logic and having edge loops. Another blog about this topic can be found here http://blog.digitaltutors.com/ngons-triangles-bad/

But I also read this Why do game engines convert models to triangles compared to keeping it as four side polygon which says that in the end everything is converted to triangles. So I would assume that it's better to use triangles from the beginning so that the engine doesn't need to convert quads to triangles.

And now I don't know if it's better to use triangles or quads.

I could think of it this way:

  • For modeling use quads, because it's easier to manipulate a quad-mesh.
  • When modeling is done convert to triangles and export the mesh. But I think this is altering the low poly art style.
  • Use in a Game Engine.

If I read the ngons-triangles-bad blog post it says that there could be render issues with triangles. So why do engines than convert the mesh to triangles?!

Am I missing something important?

Any advice or more info? Thanks!

\$\endgroup\$
1
  • \$\begingroup\$ so that the engine doesn't need to convert quads to triangles. triangulating a mesh is simply storing 6 vertex-indices instead of 4, you can do this while loading without any performance-loss. Also, many 3D modeling softwares allow you to triangulate the mesh when you export it. \$\endgroup\$
    – tkausl
    Commented Jul 21, 2015 at 16:40

1 Answer 1

8
\$\begingroup\$

A mesh should consist of what it needs to consist of.

Certain modelling operations are made significantly easier to implement, or provide better visual results, when the mesh is represented as quads. Your linked article does a decent job of provided a brief survey, and also touches on the edge-loop benefits which are useful even in simple modelling.

However, modern graphics cards only work with triangles, so at some point the mesh data must be converted to triangles.

Consequently, a reasonable rule of thumb is model using quads (particularly if your art pipeline is going to be making extensive use of algorithms that are better-suited to quads), and convert to triangles after the stage in the pipeline where the modelling is done, for example during the export/bake process that turns meshes into the optimal engine format.

If you're not doing anything in your art pipeline that benefits from quads, and you prefer triangle modelling for some reason, that's fine too.

\$\endgroup\$
1
  • \$\begingroup\$ There's also subtle difference between normals generated before or after triangulation. Generally you would prefer to calculate normals before, using quad/polygonal adjustment data. \$\endgroup\$
    – Kromster
    Commented May 18, 2018 at 20:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .