Home Technical Talk
The BRAWL² Tournament Challenge has been announced!

It starts May 12, and ends Sept 12. Let's see what you got!

https://polycount.com/discussion/237047/the-brawl²-tournament

Confused about mixing quads and tris in retopology

node
Hey everyone,  

I’m a bit stuck with the retopology of a model I’m working on. So far, I’ve always kept my retopology fully quad-based. But recently I’ve been looking at works from pro artists on ArtStation, and I see that they often mix in a lot of tris/quads. Especially in dense areas, I notice many faces being built with several triangles flowing into a single vertex, and I don’t really understand how or why this is done.  

I attached two versions of the same model with different topologies. Both are quad/tri based, but in one of them I tried adding extra triangulation (a technique I’m not familiar with at all).  

I’d really appreciate your thoughts: what’s the trick behind using tris/quads like this? Is there any rule about how many edges can come out of a single vertex? When is this type of topology useful, and when should it be avoided?  

Thanks a lot!

Replies

  • poopipe
    Offline / Send Message
    poopipe grand marshal polycounter
    Depends entirely on what the model is for.

    Neither of those are ideal for realtime rendering but the second approach is unlikely to upset anyone

    If you're making the source mesh for a bake - nobody cares except the next person to work on it .
  • scottycharly
    Offline / Send Message
    scottycharly polycounter lvl 11
    As an environment artist, I use plenty of triangles in my work. If your mesh is to be used by another professional down the production pipeline, you have to validate your retopology with them. So, a rigger might impose quand base retopology to character artists. Otherwise the in engine result is your guide. It's generally easyer to modify a mesh with a clean and mostly quad based topology. Also you have to be aware that, down the line, rendered mesh are always triangulated and if your baking or texturing software and your game engine don't triangulate your quads the same way, you end up with normal or UV glitches.
  • Noren
    Offline / Send Message
    Noren greentooth
    One thing is that what happens on a flat surface that's separated by a smoothing group or reigned in by a support loop or custom vertex normals basically doesn't matter for the shading.

    And then, the local detail you added along the longer edges of the indentations seems only partly necessary and the way you terminated it in the upper example makes no sense to me. Certainly not for a lowpoly, but even for SubD modeling. Terminating loops matters if they would go on and on, but on a small model like this, it would be cleaner to just have them continue, especially between the two indentations and since you added those vertical loops.

    It's generally said that you don't want too long thin triangles, but it probably doesn't matter in a small object like this. 
    Also, it's worth considering if shallow details like this need to be in the mesh at all.
  • DEVLiminal
    As an environment artist, I use plenty of triangles in my work. If your mesh is to be used by another professional down the production pipeline, you have to validate your retopology with them. So, a rigger might impose quand base retopology to character artists. Otherwise the in engine result is your guide. It's generally easyer to modify a mesh with a clean and mostly quad based topology. Also you have to be aware that, down the line, rendered mesh are always triangulated and if your baking or texturing software and your game engine don't triangulate your quads the same way, you end up with normal or UV glitches.
    Thank you very much for your answer. From what I understand, at least for prop models like this, there isn’t really a universally ‘ideal’ topology approach. Usually, before bringing my models into the engine, I export and use the triangulated mesh from Substance Painter, mainly because of some issues I’ve encountered in the past. Based on your advice, I think I should focus on testing my retopology results directly in the engine and reinforce my understanding through trial and error.
  • DEVLiminal
    Noren said:
    One thing is that what happens on a flat surface that's separated by a smoothing group or reigned in by a support loop or custom vertex normals basically doesn't matter for the shading.

    And then, the local detail you added along the longer edges of the indentations seems only partly necessary and the way you terminated it in the upper example makes no sense to me. Certainly not for a lowpoly, but even for SubD modeling. Terminating loops matters if they would go on and on, but on a small model like this, it would be cleaner to just have them continue, especially between the two indentations and since you added those vertical loops.

    It's generally said that you don't want too long thin triangles, but it probably doesn't matter in a small object like this. 
    Also, it's worth considering if shallow details like this need to be in the mesh at all.

    Thank you very much for your feedback. It really helped me to see my mistakes and reconsider my approach. I’ll revise the topology of this piece based on your suggestions.

    As you mentioned, I tested the second example with the long thin triangles to see if they would cause any issues. While they didn’t seem to create problems once textured, I did notice some traces showing up on the normal map. Because of that, whenever I do have to introduce triangles, I try to avoid connecting them to vertices that are too far away.

    You’re also right about the extra mesh I added for the sculpt surface detail. To better understand this and gain more experience, I plan to bake two different versions of the model — one with that extra detail in the geometry, and one without it — and compare the results.

    Thanks again for your insight.

Sign In or Register to comment.