why is this “abysmally” bad topology?

Question

So I was asking for some advice and I got this:

retopo this immediately, its abysmal. Use low poly to drive a subdiv or multires, and add geometry as needed. It's literally unworkable at this stage.

Of course I now know that I shouldn't have started with a UV sphere but it was kinda too late and only really cause some (invisible) problem on the top.

It's my first model of this kind, I only have 216h of blender (thx steam for the counter). The vertex painting is of course horrible and it's just a visual guide and will be deleted ASAP.

What my friendly reviewer didn't tell me is "why?" and, if I "retopo" the model (I won't, I'll restart from scratch), what must be done in order to not be a disaster. I tried hard and spend many hours trying to keep the quads "clean" instead of hacking my way with the sculpt tool and dynatopo. If I just wanted to sculpt I would have used oculus medium, but I want to learn blender too.

I wasn't quite ready for such a hard comment but I got over it (it was yesterday). My problem is, if I had to do it again, I would do it the same way (I would start with a quad sphere instead of a UV sphere but that's all) and it's apparently the wrong way. And I can't imagine any other way to do it.

I understand also it has too many polygons. I started with a lower resolution but once the general shape was complete I applied the subsurface modifier and started refining the model and this is what's shown in the picture. I shouldn't have applied the subsurface but I'm not sure it's what makes it "abysmally bad". Therefore, what's wrong with the topology and how to not repeat the same mistakes, whatever they are ?

Thank you.

EDIT : I'm asking this because I'm stuck in my process now. Because I know I'm doing something incorrectly but I don't know why it's incorrect.

Answer 1

Your problem is that even though you used all quads, they do not flow with the muscles underneath and they do nothing to describe the form other than define a surface. You need to watch some of the tutorials on retopology and you will see how they create loops around the eyes, mouth etc. and join them to create a mesh that is easy to animate with rigging, as well as low poly enough to properly uv unwrap and texture. High resolution sculpts can be used to derive the form and the lower poly mesh is what you animate and bake the details to.

Sorry you got such non constructive criticism, but that seems to be the norm on the internet. I would suggest taking a look over at some tutorials on CGCookie.

Answer 2

Building off of Craig's answer, you have a lot of wasted geometry in you mesh that is simply not describing anything. You could probably get away with a fifth of the geometry and still keep the model looking identical. While quads are a good rule of thumb (usually to aid subdivision surface operations), what you are truly gunning for is to make each piece of geometry descriptive and necessary. You know you are close if you can't remove any edges anymore.

Polycount has a section on their wiki for good topology, but I will warn you that "good topology" is also as much personal preference and changes based on the project.

Answer 3

Two reasons

1. Your topology completely wastes masses of polygons near the top of the head because of using a UV sphere instead of subdividing a primitive shape, subdividing a primitive has the benefit that it adds polygonal density uniformly, i.e. When you subdivide, the front of the face gets as many more polygons as the top and back of the head, whereas when you add rings to a UV sphere, you add hundreds or thousands of polygons to the top and bottom of the head and very few to the front, back, and sides, where you want almost all of your detail to be. Retopo will make make the polgyonal density as uniform as it can if you change your mind about starting over.
2. Your topology should match lines where you're planning to add rigging points to the face, usually the most significant musculature under the skin to prevent strange clipping and warping effects when the model has to be rigged for animation. Starting over will make this easier than doing a retopo, but will waste your work so far, it's a matter of how much time you'll have to put in reworking your retopo'd mesh vs how much time you'll take starting with that principle in mind that is most important to the decision.

Answer 4

I wouldn't take it too personally; you're right to want a better description of the problem, though. Craig's answer covers the long and short of the problem, but if you want a heuristic to go by consider this.

First, every vertex is effectively a guide point for the mesh, so if the vertex/triangle/quad is not contributing, it serves no purpose and makes manipulations more intensive for both sculptor/modeler and processors. Second, speaking of guides, note that under many circumstances vertex count cost can go up geometrically or worse. This is why much interactive entertainment relies heavily on LOD pruning.

As to whether your model is "unworkable", that depends on what you're trying to do with it, but a hair short of 6,000 vertices is almost certainly going to be a problem down the road. I don't know if you explicitly need to start from scratch, there are probably a dozen different ways to retopo; but if you are, for extrusion modeling I would suggest (for the above reasons) starting with something simple like a cube, and learning to use uncommitted modifiers for your details until the last possible second. If it isn't broke, don't fix it with another quad.

"A designer knows he has achieved perfection not when there is nothing left to add, but when there is nothing left to take away." — Antoine de Saint-Exupery