So: why are triangles bad when poly modeling? This is a common question posed to me. And if you’ve watched any of my videos, you know I use tris from time to time. But a lot of people will tell you they should always be avoided. What’s really going on here?
Why are triangles bad compared to quads?
Talk to a graphics programmer sometime and try asking for a poly budget for your game art – they will often give you an answer in verts, not quads, or even tris.
The truth is, computers think in terms of verts. Quads and triangles are there for us – the artist!
Quads are for the artist’s convenience
Quads are easier to see and understand. Our eyes can make sense of parallel and perpendicular rows very fast. And modern modeling tool sets take advantage of all those parallel rows of edges to do neat stuff.
But to a computer, there is no such thing as a quad. There are just two connected triangles, and the software simply doesn’t draw the interior edge. The edge is invisible to us, but there is still an edge there!
Triangles are convenient, too
A completely quadrangulated mesh is easier to take in at a glance. Conversely, all the zigzag lines of a fully triangulated mesh make it much more difficult to read. So quads have some obvious benefits.
But making a completely quadded mesh can be difficult and tedious. Triangles are usually a quicker way of resolving that kind of problem. The occasional tri can be a huge time saver!
But – triangles can cause problems in certain circumstances.
So how do we know when it’s OK to use them? Why are triangles bad? Here are rules of thumb that I use to judge when it’s appropriate:
QUADS
USE when…
- you can. These are the default! Use them when possible.
- you need to adjust large areas at a time – edge ring and edge loop tools only work on quads.
- you need especially predictable smoothing results on your mesh.
AVOID when…
- the amount of effort it would take to retopologize an area is not worth the time it would take. Just use a triangle!
- you need to control exact edge placement!
- poly-budget is a big issue, like on a mobile platform.
TRIANGLES
USE when…
- they are the easiest solution to a retopologizing problem.
- the triangle is in an area that doesn’t deform much – like areas between joints on a character model.
- you’re modeling a hard surface object – if the mesh doesn’t deform, triangles aren’t a big issue!
- you want to cut off an edge ring or loop – e.g., selecting a ring on your character’s leg goes all the way up into the back, so that you can’t focus on a specific area of the model. Strategically placing a triangle will stop that edge ring selection.
AVOID when…
- the triangle is in a problem deformation area, like a character model’s shoulder.
- your model will be deformed, in general – whether via bones or other modifiers.
- the triangle causes pinching when the mesh is smoothed.
- you are sculpting… Even though triangles work fine in Mudbox, there are some edge cases with mesh editing tools where triangles cause a problem.
Quads vs Tris – what’s the verdict?
Use whatever works. Quads are the default. But feel free to use triangles when you need to use them. There’s no hard and fast reason to say that you should never use a triangle. Sometimes it’s just easier!
Jus stumbled upon this. Great! It just made click while I retopoed my latest character and I realize, that this somewhat odd rule of “jus quads” always prevented me from making topological progress. Supporting the silhouette with a few efficiently and well placed triangles here and there can save a huge amount of additional unneeded polys. This was one of the greatest moments in my learning progress. Now I am able to get that low poly, tightly silhouette supporting geometry I was always aiming for, letting the normalmap do the rest. Great to see this written here. Good article. There is way too much misinformation on this. Especially when it comes to modeling for games.