Need opinion on what a good topology is

EndlessTinker

Hi guys, I'm new here as well as getting into the 3D content creation for games. I have been into some basic tutorials, videos, articles, and they keep mentioning about good topology, especially when it's related to real-time rendering such as games, and how important it is to keep topology clean right from the start or at least aware of it. I also look at people's work and see how good their mesh topology is (or at least I assume they are).

So along the way I start wondering, where do I start to make this well? will I end up having too many edges? Am I putting the right edges? Do this have a good flow? and other more questions came up, so I'm kinda in doubt with close to zero knowledge worrying to step further, so I need more experienced opinions to guide before I keep making my own assumptions on what's "right" in terms of "clean topology" and make meshes like time bombs waiting to happen as I'm starting to refine them further and further.

To keep this thread specific, below is one part of a simple mesh I created as an example (it's a ramp corner without fences of an arena):
 

From what I see here, there are some stretches on the side and it's kinda extreme, but I need it to be straight for the left side. Not sure how it should be.
Then there's this bevel with 4 segments on the bottom left corner, producing a non-quad face on its end somehow. Should I create more edges to accommodate this? Can I just leave it alone? If I can't, which direction should I let it flow to? Or the bevel itself is already a mistake?
Lastly, is this even enough? do I need to add more edges to make a better flow? Do I even need a better flow in this case?

ps: I kinda want to provide the obj file but seems like I'm not allowed to upload it, so I hope the picture is good enough information.

Thank you!

GlowingPotato

Hi,

When topology is the subject, we must have in mind if that asset will be animated/simulated or not. But this is valid for the final asset (the asset to be used in the final render/game).

When I'm blocking something, I don't care at all with topology or nice bevels. I don't really care for N-gons either.

When modelling a high poly mesh, I do take some care with things that might get beveled wrong, or a messy selection loop. Is very common to have shading issues in the high poly mesh due bad tessellation caused by bad topology. 99.9% quads.

When modelling the final asset (for games i'm referring to the game mesh), I do care a lot about topology for two reasons. First, if the asset will be animated, the topology must have nice flows to better displays mesh deformations. Second, if I gonna bake maps like a normal map, in general a  good topology leads to less shading erros and issues.

About your model, I can point a few things, but there are more capable people in this forum.
If you want to start a high poly from this mesh, try making your divisions move even.
I can see that you are using blender, so I recommend applying a smooth modifier, and playing with the mesh in edit mode. You will visualize how change in topology affects the high resolution model of yours.

EndlessTinker

Ah right I forgot about that, the model will have no animation (static). Thanks for the valuable input, GlowingPotato! Especially the flow for mesh deformation part.

FrankPolygon

@EndlessTinker

What's right depends on what the model is being used for, what the desired result is and any technical limitations or special considerations that need to be made for it work in the engine. There's overlap in the general strategy of laying out topology for models but what's acceptable in a low poly model will be slightly different from what's acceptable in a high poly model.

Starting with a clean block out and base mesh that has good edge flow will definitely make additional modeling operations easier. When a model is in process it makes sense to optimize as you go but avoid wasting time trying to make something that's perfect before the shapes and details are finalized. There's a balance.

Topology strategies for creating high poly subdivision models for baking are focused more on controlling the smoothing with support loops while minimizing smoothing artifacts and keeping the cage mesh organized and easy to edit.

Topology strategies for creating low poly in-game models are focused more on using the optimal amount of geometry to hold the key shapes while removing geometry that doesn't add to the overall shape and surface detail.

This sort of context is important because the use case, technical considerations and desired results all inform whether or not a particular topology strategy is optimal for the intended use. If the example mesh is still in process and will go through additional modeling operations then it's fine. If this is the final shape then there are a few places for further optimizations.

N-gons are fine for general modeling tasks but will need to be handled appropriately before baking normal textures and exporting the model. Automatic triangulation usually works but since there's multiple ways to triangulate any given n-gon this can lead to a situation where the triangulation order changes from bake to bake and export to export. This is why it's important to triangulate n-gons in a final mesh before baking or exporting. Here's two examples of how n-gon triangulation can vary.

Topology layouts and optimization are contextual. Whether or not a particular layout is good will depend on how the mesh will be used. In the example there's an outer ring of faces and edge loops. If this ring is being used for some kind of tiling texture pattern or some other technical consideration (such as an additional material shader) then it's a necessary component. Assuming there are no other special technical considerations then the rest of the highlighted grid topology isn't adding anything to the shape and is unnecessary.

Here's one strategy for optimizing the mesh that works with and without the outer ring of faces. There's a wide variety of ways to organize the topology on the top face of the shape so this is by no means the only way to do it. What's right depends on what other shapes need to be added and how the mesh shading behaves.

Subdivision modeling is more of an approximate process so there's different topology layout strategies that have unique benefits and drawbacks. It's up to the artist to pick the strategy that offers the best balance between efficiency and shape accuracy. The key is to limit or constrain smoothing errors to areas where they either won't impact the mesh shape or aren't visible.

Excess geometry that doesn't contribute to the overall shape isn't desirable. A high poly mesh that's easy to work with, subdivides cleanly and shades and bakes correctly is passable. Both of the topology layouts shown below subdivide without causing any shading errors so it comes down to which is easiest to work with and most efficient.

To recap:

The artistic and technical context of a model are important factors when choosing specific topology layouts and modeling strategies.

Try to use the minimum amount of geometry required to accurately represent the shapes while hitting the visual quality standard.

Starting with clean, well organized topology on a base mesh will make additional modeling operations easier.

Compare topology layout strategies and look for the trade-offs between shape accuracy and efficiency.

N-gons are useful when modeling but need to be triangulated before baking and exporting.

Learning from other artist's topology layouts is always good but it's also important to experiment with your own projects and get a feel for how much you can push the geometry when it comes to modeling and optimizing different shapes.

There's a lot of great resources on this forum so definitely check out the wiki the technical talk section:

http://wiki.polycount.com/wiki/Topology

http://wiki.polycount.com/wiki/Subdivision_Surface_Modeling

https://polycount.com/discussion/56014/how-the-f-do-i-model-this-reply-for-help-with-specific-shapes-post-attempt-before-asking/p1

https://polycount.com/discussion/107196/youre-making-me-hard-making-sense-of-hard-edges-uvs-normal-maps-and-vertex-counts

EndlessTinker

Frank thanks a lot for the insights (and the pics!) and for curating them. I didn't know what to search for so this is a good step for me.