HOME | DD

AlexanderLee1 In reply to stroggtank [2012-12-31 08:09:22 +0000 UTC]

It is nice to see that you agree with this topology approach, and thanks for the compliment -- it means a lot coming from you.

👍: 0 ⏩: 0

AlexanderLee1 In reply to Makk-V [2012-12-30 06:09:56 +0000 UTC]

In certain contexts, it is not a problem to have ngons. See the reply I posted above.

👍: 0 ⏩: 1

Makk-V In reply to AlexanderLee1 [2012-12-30 11:27:09 +0000 UTC]

i always learn, to have more 4 sides or ngons is very bad

👍: 0 ⏩: 0

AlexanderLee1 In reply to glitch-IA [2012-12-30 06:08:53 +0000 UTC]

Thanks for the compliment on the form!

Regarding the topology, I agree that you have to make sure that in areas that deform, the edges should primarily take the pattern of one set of parallel edges running in the direction of motion, and another set running perpendicularly. These two sets of edges naturally constitute a set of quads.

However, where this is not an issue -- for example the fingernail region, the tendons on the dorsal surface of the hand, and stable regions of the palm -- the ngons, triangles and poles are essential for creating small forms within a manageable poly count. In this issue I am in agreement with Steven Stahlberg, who has questioned the ideal of an all-quad mesh.
[link]

👍: 0 ⏩: 2

glitch-IA In reply to AlexanderLee1 [2013-01-01 11:08:34 +0000 UTC]

Hmm that is a really interesting but what you need to bare in mind is that the models that he has done there are smoothed and regardless of an ngon or tri, when you smooth, the 3D application will try its best to calculate it and turn into a quad. From what I know ngons can cause a few issues here and there during this operation and therefore should be avoided buuuuuut looking at the Steven Stahlberg model, I'd say he knows when to use and when not to lol I've just been told by my supervisors to simply avoid it all times as its better to be safe than sorry and hand the model down the pipeline and get the blame for a car crash at the end xD

👍: 0 ⏩: 0

Omar-Kamel In reply to AlexanderLee1 [2012-12-30 22:10:09 +0000 UTC]

I thought of Stahlberg as soon as I saw this hand. And I agree. I was beat over the head with quads when I was learning 3d, but I've found that Ngons and tris can actually be very helpful in Sub D modeling, both for hard surface and organic.

If the model is built for SubD then nobody will know (or care) whats going on in the zero level mesh. Every model has poles. They are necessary. Ngons and Tris can be looked at as poles that havent been subdivided yet.

I think going for quads should be a goal, but not to the point of wasting your time or hurting your own model. I've seen many people (including myself) try to beat a model into submission for all quads and sometimes it just makes things worse.

Sorry for the rambe! I just thought you should know someone else understands. XD

And if you look at models from Pixar, Weta, ILM, etc... all these top notch studios have Ngons riddled throughout their models.

👍: 0 ⏩: 1

AlexanderLee1 In reply to Omar-Kamel [2013-01-02 08:43:19 +0000 UTC]

Yes, poles are actually what make it possible to properly align edges with the direction of form curvature and articulation. That's a great point about ngons and tris being unsubdivided poles, I hadn't thought of that before.

Thanks for the comment! I like to discuss this sort of thing, because I think there is a lot to be discovered about the best way to do topology. You know a good forum thread where I could discuss this subject (looks like Stahlberg's isn't active anymore)?

👍: 0 ⏩: 2

Moybella In reply to AlexanderLee1 [2014-06-24 01:51:53 +0000 UTC]

There are a number of reasons why quads are considered better practise,  its not just about topology, but deformation (which you covered), uv's and rendering, some renderers deal better with Ngones then others, like Renderman, then you have your subD method, OpenSubdiv is much better at smoothing ngones and poles, hence why coupled with PR Renderman, the pixar lads don't mind being cowboy's when it comes to poles and topology. as for ILM and WETA they may have they're own reasons i'm sure, but for instance, when a character is going to be completely covered with fur or elephant like hide, it does not matter as much.

Displacement would be another consideration tying back to clean UV's

some game engines don't support ngons or tries outside of tessellated polygons.

Often though it is the industry peripheral software that require this, like mudbox and zbrush. programs that are used for a multitude of tasks , previz blendshapes... etc often these get integrated fairly heavily into the pipe and you have to be able to back and forth a mesh between multiple departments.

at the end of the day it also depends a lot on what you intend the mesh to do.

however i would caution you on demoing such topology in job interviews, most companies won't get it.

I'm actually quite interested in this discussion, does this Stahlberg have any other resources ?

👍: 0 ⏩: 0

Omar-Kamel In reply to AlexanderLee1 [2013-01-03 02:02:37 +0000 UTC]

It is a very interesting subject. I was obsessed with topology study for some time. I cant really think of any discussions going on now though. There used to be the thread at the SubD Modeling site before they closed down and then the one over at CGTalk, but I dont think its active anymore.

If you happen to come across one, or maybe start one yourself, let me know! I always enjoy seeing the many different topolgy solutions people come up with. Its an art form unto itself.

👍: 0 ⏩: 0