Some basic questions about retopology

Support Modeling
1

Hi Guys, I am currently learning the basics of retopology. I understand it as recreating a low-poly mesh basing on a high-poly one and then apply a normal map generated from the high poly mesh to the low poly mesh. (please correct me if I was wrong xD)

So my questions are:

  1. when do I use multiresolution modifier? is it necessary? I am aware that it can be used to smoothen and adding details to the low poly mesh. However i do not find it any different from the subdivision surface modifier. Also it does not support dynamic topology so I assume not many details can be added with this tool.
  2. I came across an article saying in retopology for majority of the 3d modeling every individual face should have 4 edges and connect to 4 vertices(a quadrangle? ). In gaming, a triangle face is mostly used. So what does this mean and why is this? In my personal experience so far, I have been creating single faces with a range of different edges/vertices and I do not really see any problem.

maybe the questions are a bit unorganized :blush: / Any help would be much appreciated!!!

2

As far as the quads go in the modeling is that they tend to deform better and offer a better smoother animation to the object with quads, also subdivision surface doesn’t really like triangles very much. Which is why most characters are retopologized with a smooth looping flow in areas that require the most movement during animation and posing/rigging. As for the multires I don’t typically use it that often I tend to just get away with a subsurf and dynotopo for sculpting. But I am sure others can chime in on when and why and how to use it.

1 Like
3

You can use the multiresolution modifier for sculpting purposes, with with dynamic topology available, you don’t really have to use this as much anymore, but someone with more sculpting experience could answer this better.

Modelling in quads only is a good practice, however not always possible or required. It is best then to keep in mind that you should TRY to model in quads only. Quads work better for deformation, edge/poly flow, makes editing easier especially when using subdivision, good to use in organic models.

Under the hood, everything is triangulated, you just don’t see it. From what I understand though, if you model in quads, (for the sake of your sanity and ease of editing) once you bring the asset into a game engine, it will be triangulated automatically, I don’t think any of the game engines need pre-triangulated meshes anymore, but I could be wrong. I believe that the reason for the tris only is because triangles are always planar and rendering requires triangles for ray intersection. Quads and shapes with more sides can be non-planar and cause a lot of issues.

1 Like
4

Ok, I’ll chime in even though I don’t consider myself to be an experienced sculptor.

Dynotopo is the newer sculpting method added to blender. It dynamically tessellates the areas of your mesh you are sculpting on, increasing the resolution of those parts as necessary. The downside is that this is always triangulated geometry and isn’t suitable for uses aside from static sculptures.

We stick with quads in part because it makes it easier to place loopcuts across faces, and we need that sort of control when placing extra loops around areas that deform. Tris can also cause pinching when using subdivision surfaces because catmull-clark converts tris to quads.

All polygons are converted to tris before they are rendered, even in cycles (Test out the wireframe node there). Some game engines can’t handle the conversion from quads to tris on their own for some reason. This is why game asset makers sometimes have to triangulate their meshes before export to game engines.

Sculpting also becomes less responsive sooner when using dynotopo than with the multires modifier. They say when you split your mesh into sections, you can sculpt around 1.6 millions polygons on each section with dynotop before slowdown. With the multires, the limit is supposed to be a lot higher (around 10 million polygons).

Multires is optimized for the high poly to low poly workflow. You sub-divide your base mesh with the multires modifier, and sculpt on the highest detail level. When you are done, you copy your mesh, set the detail level to one of the lower stages, and apply the modifier on the copy. You’ll have a low poly mesh that more closely matches the highres version that you can bake the highres details onto.

The multires Is also far less destructive, you can add more detail to a quad based mesh and end up with a quad based result that doesn’t need retopologizing.

Getting predictable results also allows you to use it for tricks like this:


Dynotopo would cause the number of verts to be different between the two meshes.

The one downside to the multires modifier is a long standing glitch where sculpting on a lower subdivision level can cause artifacts in the higher levels. The modifier is still perfectly suitable for sculpting if you only work on the highest level though.

2 Likes
5

All good points. Thanks for “chiming in”.

1 Like
6

This answer is truely helpful!!! I now understand the purpose of using quads much better.

Also please allow me to ask another question: I find that sometimes when applying a normal map to a low poly mesh there are obvious low polyed areas remaining (It seems like the normal map does not cover the entire mesh so the low polyed base pops out). Would you say this is resulted by the insufficient number of quads I use?

7

Thanks so much for your help! Do you create the base model using subsurf and dynotopo and then just conduct retopology base on that?
Currently a problem im facing is that the normal I baked from the high poly mesh done by dynopo tend not to reserve all its original detail textures. When applying this normal map to the low poly mesh not everything is covered (the low polyed base show through the normal map on some areas). This is why I am assuming i should try to use as many quads as possible to avoid it. Or would you see any other potential cause of this problem?

Sorry for an additional question xD much appreciated again for your answer to my previous questions!

8

Great points there! Thanks so much for your help. I truely feel what you said about “high poly to low poly work flows”. I have been applying multires to many simple mesh i create currently. However for meshes that expect to have a large amount of details i still cannot help to use dynopo to do some tweaking and adjusting at this stage. Maybe getting used to multires will grow me into a more decisive sculptor too!!

9

What does the actual normal map look like once you bake it? Is there solid colors in it that seem out of place? Or do they look like a typical normal map?

10

This sounds like an issue with the ray distance setting. You have to adjust it inorder for blender to see the topology in certain parts of the highpoly mesh.

Ray distance set to 0:

1
11920Ă—1036 411 KB

Ray distance set to .5:

2
21920Ă—1036 364 KB

Here’s the blend file:example.blend (1.2 MB)

11

oh great ! thanks for the file. I just tried it out i think this solves the problem. However, for some complicated mesh i feel like ray distance is very hard to control (it is difficult to find a appropriate number/ do i have to try out each number individually?maybe it is a experience thing?)

12

As far as normal map experience, I’m not much better at that than you are. I mostly just do hard surface modeling. I rarely make my own normal maps. Yea, I just play with the distance. You don’t usually need to turn it up too high. You typically only need to turn it up to .1 or .2 in most of the cases I’ve tried.

13

Sure, I’ll definitely look into them. Thanks again!