When are triangles acceptable over quads/squares?

[ThE_JacO]

Actually animation is the one thing where you could triangulate every single quad in a mesh, and nothing would change -at all-.
The whole thing that all quads should always be compulsory is frankly an out-dated and largely mis-interpreted notion. From a technical stand-point, particularly for tool compliance and rendering, there are advantages to quads (easily identified loops, good response to spring networks for various tools that build one out of topology, good response to subdivision etc).

There are other things, however, that people obsessed with the rule but ignorant of its implications do to keep an all quads topology which are way worse than having a three or five sided poly in the topology.
Excessive amounts of topological singularities, non convex polys and other things often forced on those poor models will mess other things up and cost blood down the line.

The subject has been discussed a million times before. Use the search engine first, and once you find the threads engage your brain and do some research, since there's a ton of misconceptions floating around about the subject.

[Wiro]

I'm moving this into the modelling forum.

As Raf says, one can be too obsessive about all-quads. What you have to look out for is how the geometry changes when smoothed (subdivided) and whether it becomes more visible when deformed or not. If it still looks smooth, it's fine.
Other than that you have to know if something that will read your model expects quads. For example ZBrush can act up with tris.

Wiro

[Leionaaad]

I am one guy who is actually obsessed with quads. The reasons are all mentioned above: predictable edge loops, no artifacts at all at smoothing/subdividing, skinning is a lot better with quads, and the list can go on and on.
The fact is once you get the method to model that way, is not hard at all. The surface will be somewhat heavier, of course.
The other thing is: sometimes I broke the rules, but I always go with triangles, never with n sided polygons, I always put these things(like triangles, pinches) away from sight/deformation

[ThE_JacO]

skinning is a lot better with quads, and the list can go on and on.
Sorry, but the list can't go on and on I'm afraid :)
How is traditional skinning, which only displaces vertices and displays a mesh with polys divided in triangles for drawing anyway, affected by the number of edges?(which like it or not is the SOLE difference between all quads and all triangles).
Explicitly split all your quads in triangles the same way OGL will do, and tell me there is ANY difference between that and what you had before.

All quads does NOT benefit skinning in any way. The sooner people get that out of their heads, the better.

[Leionaaad]

You're right, I haven't worked on messed up meshes for a long time. You are absolutelly right with the triangles being skinned right. However, I remember I had hard time painting weights on n-sided polys. Once I traced one edge it was better. (I didn't remember wich annoyed me: triangles or n-sided polys) the thing with these issue always boils down how it is subdivided, anyway: not a good idea to have pinches where the geometry has complex curvature or it deforms.

[errantspark]

The quads over tris rule is generally a good one, but rules are meant to be broken, there will always be exceptions where a tri or a five sided poly is more useful. Keep in mind though theres a rule about breaking rules, you have to know the rule to break it. You have to understand why it's better to use quads before you can go throwing tris and n-sided polygons into your models and expecting everything to be shiny.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.