So I just started learning how to model, and before I go any further I’d like some opinions on what the best way to lay out edges in a mesh is. I just started modelling a low-poly building for a game I’m making (I’m a programmer, hence why this is a topology question rather than anything artistic ;)), but I don’t know which is best out of these two approaches (see the attached pictures).
The first one (triangles.jpg), is just what naturally happened whilst getting the shape I wanted (by subdividing, extruding, moving, merging). The second one (quads.jpg), is what I got after I rearranged things by doing more subdividing, rotating edges and more merging. To me, it seems the first is good in two ways; it took less effort and it has a lower poly count. The second seems better because it just looks cleaner, and I can do things like selecting edge loops more sensibly, which I could see being useful once the model gets more complicated. I think the second approach is probably better.
So which approach do you nice people think is better, and perhaps more importantly am I reasoning about this in the right way?
(also ignore all the other bad modelling in the screenshots, I’m just mucking around at the mo. Concentrate on the differences).
Cleanliness won’t matter as long as you don’t see it.
Anyways, all quads will be converted to tries when saving runtime, or even in most other engines. But how I get low poly game models is that I loop cut, subdivide, all I need, and go back and merge unneeded faces and verts.
Most people find that modeling in quads is faster, since there are a lot of tools that work well with quads that don’t work well with tris. If you model faster with tris, do it. As Eels mentioned, game models get converted to tris anyway.
Your observation about more complex models is spot on. Since the lastest techniques involve making a high poly detailed model, then burning a normal map to put on a low poly maquette, you’ll wind up doing a lot of complex high poly modeling. In those situations, modeling with quads winds up being the timesaver.
I know the quads are actually two tris, but obviously in the first shot there are three tris on the rectangular bit whereas in the second shot there are two quads = four tris, so modelling with quads seems more expensive to me. And the side of the building that’s facing to the left is just one quad in the first shot, but two in the second (aside: is it normal to just insert an edge loop all the way round something to add detail to just one area? Won’t you end up with a lot of unnecessary edges in the middle of flat faces?).
That said, it seems like it wouldn’t be a hugely difficult problem to algorithmically reduce the poly count of a mesh when there are so many coplanar triangles with shared edges. Are there any tools or options in exporters to do that kind of thing? If so, the decision becomes very easy.
A quad is not simply two tris – subsurf a mesh and you’ll notice the difference (I come from AutoCAD, and it took me a long while to accept this).
When modelling buildings, I typically don’t want to have diagonal edges across facades because it makes it hard to add windows etc. But as long as you are not planning to animate your building (e.g. tear it down) I don’t think it matter if you use tris or quads.
That’s confusing What happens when the vertices of a quad aren’t coplanar? How would Blender render that? And presumably on export all quads will become two tris, or am I assuming too much again?
What happens to quads on export depends on what you are exporting to, I guess.
Here is a badly non-coplanar quad with the subsurf modifier applied to it. The quad looks more like tetrahedra. Of course, if you model things really badly you end up with render artefacts (been there, done that).
Is there meant to be an image in that post? It’s not loading for me.
If I’m exporting to a triangle mesh for use in a game, for example using the Ogre exporter, that will presumably convert all quads to two tris, right?
Also going back to this: “aside: is it normal to just insert an edge loop all the way round something to add detail to just one area? Won’t you end up with a lot of unnecessary edges in the middle of flat faces?” - is that the case? None of the tutorials I’ve seen mention this kind of thing, they’re all very specific to the thing they’re trying to model and don’t give much general advice like this.
When using subsurf you typically add extra edges where you want edges to be sharp. To add more detail in general I know nothing better than plain old modelling.
Ah thanks, I see what you mean now. I don’t know how a non-coplanar quad will look when exported for a game, but if I were to do what Star Ranger4 does and convert to tris before export this wouldn’t be a problem.
What happens when the vertices of a quad aren’t coplanar? How would Blender render that? And presumably on export all quads will become two tris, or am I assuming too much again?