Home Technical Talk

Holding edges (bevel) while maintaining quad based geometry

Hello everyone,
this is my first post here, so please bear with me. I am right now trying to add holding edges to my model for subdivision. I'm doing this by beveling the edges and manually cutting the polygons to maintain a quad based geometry. Since my method produces a bunch more edge loops and thus more polygons, I am looking for a smart alternative.
Maybe one of you guys has an idea to go about this.
(The images show how I started.)


  • Kanni3d
    Offline / Send Message
    Kanni3d ngon master
    It's for sub-division as you said, therefore polygon count doesn't matter.  It's only use is to give you high-res beveled edge information, and wouldn't be needed for much else.

    You'll end up duplicating the mesh and deleting all those support/holding edges as they aren't needed for your in-game mesh.
  • Elytreus
    Thanks for your answer. In short, my approach wasn't wrong for the current use case, since I have to avoid deformation.  If I wanted to use the model in a game and it exceeded the polygon limit, I could just delete the edges I added to achieve a quad only mesh, since it would be triangulated anyways, right?
  • sprunghunt
    Offline / Send Message
    sprunghunt polycounter
    I always meshsmooth/polysmooth hipoly models before I export them. This will make sure that there are only Quads irrespective of what your actual geometry is. It does mean you need to build your models so they smooth correctly.

    Something to note is that the N-gons in your model will probably be a bigger problem than any triangles.

    And lowpoly geometry is fine to have triangles.
  • Nominous
    Offline / Send Message
    Nominous polycounter lvl 9
    You seem to be missing support loops outside of your bevels in order to control their softness. Without these, the large, flat areas will look kinda blobby:

    On flat surfaces, you can terminate support edges early instead of running them all across the model. This can cause pinching depending on how close surrounding details are, so you'll have to experiment. This means that you don't need all-quads on flat surfaces and you can get away with n-gons and triangles. This doesn't work well for complex curvature since it can result in undesirable pinching.

    A disadvantage of terminating support edges early is possibly missing pinching until you think you're ready to bake and end up having to fix the high poly. It's always an option though. Modeling in all or mostly all-quads makes subdivision more predictable and lets you edit the base mesh easier using tools like loop/ring select.
  • Elytreus
    Thanks you all for the advice. I'm currently going over the model to fix it and will post an update as soon as I can.
  • FrankPolygon
    Offline / Send Message
    FrankPolygon godlike master sticky
    @Elytreus Is there a specific technical limitation that requires the mesh be all quads? As Nominous mentioned: Support loops will hold the shape transitions along the edges and flat surfaces are generally uneffected by non-quad topology. Subdivision modeling tends to be more approximate than exact so more efficient topology strategies can produce comparable results in less time.

    Here's an example of a quad grid block-out where the support loops were added with a chamfer / bevel operation. Subdivision smoothing rounds the edges and the width of the support loop controls the size of the round over. Connecting the corner vertices around the support loop maintains all quad geometry without having to manually add additional edge loops.

    Compared to adding in actual rounded edge geometry: this strategy takes less time and uses less geometry which means it's easier to edit in the future. Excess edge loops could be removed at the block out stage for an even cleaner mesh.

    Here's an example of quad grid topology with rounded edge geometry. All of the necessary support loops are added in a single operation by selecting the entire loop for each of the outside edges and running a chamfer / bevel operation. Rounded edges are added by selecting the outside edge loops and running a chamfer / bevel operation.  All quad topology is maintained by connecting the corner vertices between the support loops and the round overs.

    Since this strategy doesn't require manually adding support loops it's relative time efficient. The additional geometry used to round the edges increases the shape accuracy but this additional geometry can make the mesh difficult to edit later. Starting with a larger quad grid would further simplify the mesh and make it easier to work with.

    Here's an example of n-gon topology with rounded edge geometry. Shape support loops are added by selecting the outside dges and running a chamfer / bevel operation. Outside edges are rounded over with an additional chamfer / bevel operation. Support loops are added to provide structure and n-gons are triangulated near critical corner locations. Minor loop discrepancies can be flatted if desired.

    Compared to other topology strategies: using n-gons and triangles on flat surfaces reduces the amount of superfluous geometry which increases edit-ability and time efficiency without sacrificing any significant visual accuracy.

    Here's a comparison of the subdivision previews and the various topology strategies. Like Eric mentioned: Unless there's a specific technical limitation or a mesh smoothing error, topology on flat surfaces is largely irreverent.

    There's no real advantage to maintaining all quad geometry just for the sake of maintaining all quad geometry. Using too much geometry and chasing perfect topology flow on areas where it has minimal impact can become a huge time sink with minimal quality improvements. When it comes down to it, subdivision modeling is about making trade-offs between accuracy and efficiency. Optimize wherever possible.

    To recap:
    • Block out the basic shapes while maintaining clean topology with minimal geometry.
    • Use automated tools and avoid excess manual work whenever possible.
    • Use flat surfaces to terminate excess edge loops.

    There's a wide variety of modeling and topology strategies so it's worth taking the time to test and evaluate what works best for a particular shape or project. Experiment with trade-offs between shape accuracy and time efficiency to learn how to maximize the return on time spent.

    A great resource for this topic is the how do you model this sticky:
  • Elytreus
    Thanks again for your detailed answers. That was way more than I expected. I took your advice that I don't have to avoid triangles at all cost and remodeled the piece. Right now I am fairly happy with the result (If I didn't overlook something ^^).

Sign In or Register to comment.