Triangles and Cartoon Workflow Thread

I read/hear often "Never ever use triangles!" when it comes to modeling, as it is said they cause all sorts of problems. However, whenever I watch a low poly modeling video/tutorial, they always use triangles.

On a related note, I often read/hear the game character model workflow goes like this

-Base Mesh
-High Poly/Sculpt
-Low Poly
-Bake and Texture
-Rig

The problem I have is that I'm working with cartoon characters, so the high poly/sculpt phase seems useless. I can do the nostrils, mouth and fur maybe. Clothing wrinkles if they are wearing any, but generally there isn't anything really worth doing high poly. Particularly since I'm just starting out. I was wondering what I should do. Do the whole process, doing high poly anyway? Do I make the base and then go straight to low poly? Or should I jump straight to low poly?

There is always the question what something is done for in 3D. If you will show objects close up and with a lot of detail, you will need highpoly. If it is for (realistic/detailed) games you need to bake the detail from the high poly into the low poly with normal maps. If you are doing cartoony low-poly meshes with, for example, hand-painted textures, there is no need for and highpoly sculpting/modeling. It's all dependant on the purpose.

The hgh poly workflow makes more sense, if you think about it like this:

1. You have a rough idea, or some concept art, so you make the basemesh
2. You add more and mode detail, thus creating the highpoly from the basemesh.
2a. Your sculpted detail is too much for the original basemesh to display correctly. You have to re-build a low poly mesh that has enough detail in the places you sculpted/detailed. This can be done from the basemesh or better by retopologising the highpoly. Simply said - you take a progam like xnormal (or your own modeling application), turn on the magnet function and build a new mesh along the lines of the highpoly. then you get the highest representation of detail on a lowpoly mesh (through baked normal maps)

3.Rigging can be done to anything, even to primitive shapes. Rigging basically mean weighing the parts by painting "weight" onto each one. Then when you move one part of the object something else will be influenced by it (or not - depends on the weights).

If you don't need it, you don't have to use highpoly models. There are plenty of ways to make detail even without sculpting, e.g. by painting normal maps in Photoshop and applying them. I suggest looking more into materials and how all that works. The latter technique works well on walls and simple meshes, but characters with details usually adhere to the process you and I described.

Edited by: SinKing

Triangles are perfectly fine. I don't think anyone who tells you to never ever use them really knows what they are talking about; more likely they heard it from someone else long ago and are just perpetuating the myth. Everything gets triangulated during rendering anyway (1 quad = 2 triangles), and for this reason it can even be recommended to manually triangulate your mesh before baking a normal map to prevent the vertex normals from shifting should the renderer triangulate the mesh differently than the baker. You'll also find that triangles can be helpful when it comes to controlling edgeflow, keeping meshes optimized, and helping with deformation in really low poly meshes.

That said, keeping things quads can still help. Modern sculpting programs wont be killed by a triangle, but you usually get the most consistent brush behavior from a quad. When subdivided organic models you might notice a sort of pinching near triangles and poles (although sometimes modelers can use this to their advantage). For hard surface subdivision models triangles and n-gons are also fine when left in more planar areas. Lastly with some programs, quad-based topology can make it much easier to select edge loops and ranges. Long story short, they're all just polygons and each have their benefits. Don't avoid triangles because someone told you to, use them so that you'll learn how to use them.

As for high poly meshes, only use them if you feel you have to or if you think they will help you. There are many ways you could approach any given model. Sometimes it's nice having a higher poly model to reference, sometimes it's easier just to leave it out completely. It is worth noting that high poly sculpts/models don't always have to be high in detail or realistically modeled. There are plenty of examples of good looking cartoonish sculpts

Edited by: Cryrid

The workflow you should use is the one that you're most comfortable with. I've seen a LOT of people go from concept to low-poly to texture to game without even thinking about high-poly normal mapping. I've seen a lot games that go that route as well.

If you're not working for/with people who absolutely require high-poly normal mapped models, don't even bother with them if you don't think you need them.

What is easy to work with? 10 faces or 20. Then you have your answer toward using tri or quads.

Edited by: vfn4i83

I think the myth about triangles comes from ease of modeling. Personally I find working in quads to be less confusing. You also can more easily identify good and bad loops and poles when working in quads. I think a lot of people feel similarly. Probably someone said this once and the receiving person misunderstood it to be tris are bad where ad they are really just less convenient.

I usually work concept to base to high/texture simultaneously. I say this because generally I can get the effect I want in crazybumps. However there are times that crazy bumps acts weird so then I make a high poly sculpt of just that region and bake those normals then copy that map portion into the full map. I generally find image to normals faster than sculpt to normals but everyone has their pref.

Edited by: endie

the only reason not to use triangles is in high poly modelling and sculpting because its difficult to imagine what, when the mesh is subdivided, the triangle becomes. but it really makes no difference in low poly modelling, in fact i would encourage it. in a low poly model if you need a triangular shape dont add the extra vertex because you dont need it.

So I would say for cartoony characters, you can model straight into a low poly if you poly count needs to be low, but otherwise its still good to practice that work flow by going into sculpts. If you're comfortable making models with like mid-range poly counts (if that makes sense) then go with that before bringing them into a sculpt so you're only adding minute detail instead of trying to form the entire character.

As for the quads vs tris discussion, personally I prefer to work in quads.

Like others have said above, it's easier (at least for me) to see where I dun goofed and to add loops faster. Though a down side is that if you're using an engine that prefers or demands tris, you'll have to convert quads to tris, and your poly count isn't the same anymore, but that's more like bragging rights to me. (ie. "Oh, that model is 2k poly? Wow, I did that in 500..." )

because its difficult to imagine what, when the mesh is subdivided, the triangle becomes

3 quads

Though a down side is that if you're using an engine that prefers or demands tris, you'll have to convert quads to tris, and your poly count isn't the same anymore

Triangles are polygons too, so the count should remain the same if it were initially reported in triangles (which is ideally what people should be using when reporting triangle counts, as anything else is pretty deceiving for the artist to use). To use an analogy, "polygon" is like saying the word "coin", while saying 'triangle' is like saying '1 cent'. If someone tells you they have 30 polygons, you still don't have an idea what their mesh is like just like you couldn't really guess how much change is in their pocket if they told you they had 30 coins in there. If they tell you they have 30 cents, then you know.

The vertex count is an even better indicator of a mesh's size (as bad UVs and smoothing can dramatically increase it), and that number shouldn't change on triangulation.

Another Cyrid Puzzle for me
I need a drawing again - how does a triangle become 3 quads. I can make one quad and two triangles out of it, but I have problems with 3 quads. I have an idea, but it looks very unmathematical (two quads on top of each other and a triangle sharing the border with them, thus turning into the 3rd quad.)

Yes, for game production triangles are totally okay to use. If you have animated meshes however and a mix of triangle and quads on a very highpoly (non-realtime) model, there can be issues with materials during rendering. But that is unrelated to anything related to games, however that was something they taught me at school. I never had any problems with triangles on meshes, but my teacher sure had those kind of problems with triangles in my work. I totally hate doing annoying things that cost time for no reason, so I'm with the pro-triangle faction, now.

From what I've seen in 3D and my own work so far, it's all a process of constant optmization. Perhaps there is a point where you reach the perfect state, but until then modeling is about making things work, fast. So, if you need 3 hours and 14 new edges to turn a triangle into a quad, it simply isn't worth the effort.

Most likely the change is invisible to human eye. I believe our teacher drilled us for triangles, so we'd keep our models organized from the start. I'm not sure about it, but if he thought we'd keep doing quads where they are not needed, he's a fool. I'd say over 95% of my meshes are quads anyway. If you look at supercool models from games you will find they have a lot more triangles than my "schooled" work. By that they also have a lot fewer polygons in total and thus have better performance on the game-engine than anything my teacher would have let me make. Which shows that teachers and education in the 3D field are very overrated. There is no Axioms in 3D modeling workflows. EVERYTHING you find out is either by finding someone else was right or wrong, after you tried it for yourself, first. Do try

Edited by: SinKing

I need a drawing again - how does a triangle become 3 quads. I can make one quad and two triangles out of it, but I have problems with 3 quads. I have an idea, but it looks very unmathematical (two quads on top of each other and a triangle sharing the border with them, thus turning into the 3rd quad.

Basically n edges = n quads. Subdivide a triangle and you get 3 quads, subdivide a quad and you get 4 quads, subdivide a 5-sided polygon and you get 5 quads, etc.

Yes, lower left would have been my idea, too. I'm just getting to triangulation in my own models. I understand it makes sense to do it manually, in order to avoid errors on the normal map and detail shading.

Triangles are bad bcs when you subdivide they turn into a different shape entirely whereas w/ quads they retain their shape and proportion giving you the same mesh flow you started out w/, now just more of it. Avoid tris as much as is possible.

Re: workflow, the philosophy between doing a realistic char and cartoon char are different to the point of being antithetical. W/ a normal mapped realistic char the illusion of realism is achieved by faking geometry where there isn't any using the normal map. The silhouette is often still boxy as f*ck. With a cartoon char, the silhouette is everything and therefor you should distribute your polys slightly differently. For a cartoon character your hi-poly sculpt may not entail anything more than hitting turbosmooth.

triangle's are not bad or good generally but they are bad and good for different purposes. if you aren't going to sculpt the model for a normal map than there is no reason not to use Tris for video games. if they have to be sculpted than it may be best to wait to triangulate until after you sculpt it and have your normal map to use. Tris do not subdivide well but you shouldn't be subdividing and leaving the mesh to be imported into a game anyway. if its low poly then its most likely best to use Tris. if you know you are going to use Tris then before you import it into the engine you are using triangulate it, don't make the engine triangulate it to save computing resources.

the use of quads in games also comes down to ease of animation, try rigging a triangulated mesh

That depends on how you rig more than anything. If you need edge loop selections then quads will help. If you're just painting or selecting verts and adjusting the envelope there, 100% triangulation shouldn't be a problem. And to find the middle ground (since the question was regarding 'Never ever use triangles'), well-placed triangles can make deformation (and thus rigging) even easier (Wiki.polycount.com has some examples).