Optimal polygons for a game engine?


I originally asked this question here, but I didn’t get many responses. I decided to move it to a support topic, so I can get more help from the support community.

I am currently working on a set of characters. The plan is to eventually use them in a game engine. It’s too early to tell what engine is going to be used, but I would like to have some assets done by the time it’s developed.

I would like to do a low poly style, and though this means it will be a lot easier in some ways, it also means the shape of my polys will play a huge role in the end result.

This brings me to my question:
Should I model with tris or quads? In the end, it will be rendered as tris. Should I triangulate everything ahead of time so that I have a good idea of what it will look like and I’m not leaving everything to the computer? Or should I use quads so that they deform properly in the animations? Will n-gons be tolerated? I have little experience in this type of project and would be grateful for some help.

I’ve never really concerned myself with making game props, but I do (or at least did) have this efficiency before all else mindset when it comes to modeling, so I guess I can comment a bit.

It’s hard to nail down just how many polys would be best practice for a character, with the general consensus being anywhere between 10,000 - 100,000 for your main character, half that for NPCs, and around 500 to 20,000 tris for various props.

…seems there’s a lot of play when it comes to polycounts.

For tris, you’ll want to model and UV first in quads, then convert to tris before you texture, and export it out to a game engine. This isn’t something that’s super duper important that you ABSOLUTELY must do at all times, but having another program convert your quads to tris can occasionally introduce some shading errors on your models, and converting it yourself helps to get ahead of it.

If you’re working with static, hard surface objects with only a slight amount of curvature, ngons are fine. But for more organic objects with fluid moving surfaces that you intend to animate, like trees, plants, animals, or people, you’ll want to stick with quads.

So if you have a boxy sci-fi crate with lots of little nooks and crannies, ngons are okay. A deer galloping through a forest with all the trees swaying in the breeze? All quads.


Thanks for the advice! (I’m gonna need it…)

Here is my current prototype.

I’m doing a low poly style, so I won’t have to worry about smooth shading or number of polygons, but I will have to worry about deforming, and triangulation could drastically change the look of my model.

There are only a few triangles aside from the head, which is composed completely of tris. These are on the tip of the hat, and I’m not sure, but I believe they should be fine, and shouldn’t affect the deforming of the hat.

The head is a different story. I will probably want to rig the nose, as this will create most of my facial expressions. The eyelids probably need a few extra edge loops, but I’m not ready to apply the booleans yet.

Keep in mind that I’m not an animator, and my attempts at character modeling have been pretty basic, so any advice I can offer here is limited.

At a casual glance, I’d say the hat and body are fine for animating, but I’d retopologize and add a little more geo to the face to allow for smoother expressions without too much faceting in your shading. Your eye sockets in particular will probably cause you tons of problems when you start moving things around, with weird, tight vert groupings there in that one corner, and what looks like an ngon at the other.

If it were me doing this, I’d remove the socket altogether, and leave the eyes freefloating on top of the face. Throw in some eyebrows for some extra character.

1 Like

Quads for deforming models, what you like for the rest


you want quads for better deforming

No. And you are not gonna have any ngons if you triangulate everything before output


EDIT: actually you may want to have quads everywhere to avoid all those tris-problems (complicated unwrapping, possible mapping problems, wild normals, uneasy rendering)
anyway it’s not a rule, it depends on the model

1 Like

Quads are generally easier to view and edit. It should be noted however, that it’s not a crime to use tris here and there in an animated game model. A well-placed triangle can spare you some headache and polybudget.

1 Like

The GPU only works with triangles. The polygon within the visualizer is a triangle. This is the minimum necessary and sufficient amount to build a normal to the surface and, therefore, to build a shading. Always, even in the Blender viewport, when you create a polygon with more than three points, you create a subset of triangles and Blender approximates their normals. Sub D Modeling has been a requirement for animation since the days when two engineers from Pixar proposed this algorithm. Today, modeling only with quads is convenient for the division algorithm, it is a standard in the industry and in general it will be more convenient for you to think and work like this.

1 Like

@Renzatic, @kabu, @StrayBillie, @baldperv,
Thanks for the help guys!
I have a long way to go on my 3d journey, but I’m really enjoying it.

I have a lot of time on this particular project, but I wanted to get ready soon so that I can at least pretend I know a little bit about what I’m doing when we get devs on the project… :grin: In the end, I’ll eventually learn. That’s what deadlines are for.

I just want to make it a point that I’m still absolutely open to suggestions, tips, and advice. After all, I’m going to need it.

Thanks again to my favorite community!


I’d say use mostly quads since they are more convenient for mesh editing tools (subdivision, proper UV unwrap, etc) to work with. Keyword “mostly”. Sometimes, resolving everything to quads requires cutting an extra loop through character’s whole body; then better use a triangle to avoid it. Also, it may be a good idea to triangulate everything right before exporting into your game engine of choice (after modeling, texturing and rigging is finished), just to test if everything deforms correctly, and if there are problems, re-triangulate those problem polygons in different order. Blender and game engines work with meshes differently. In Blender, triangulation of polygons is dynamic, dependent on degree of their deformation, while game engines operate with fixed triangle face tables. https://gyazo.com/845edeff4c6179a942ccdac0f9dca402

1 Like