Body topology


Instead of keeping the SubD hierarchy, which was my old worlflow (see old tute on my site), I’m now doing what everyone else does, putting all the detail on the 0 level. (I discovered more drawbacks, and I’ve found I could get the benefits other ways.)

As you can see I’m pretty much ignoring the edge-loop concept (unless they appear by accident), and I use 5 and 3 sided polys whenever it serves. This gives a lighter model, important if you want to make lots of Blendshapes, and it tweaks more easily. It also leads to better deforming. The whole thing is only 3400 polys in polygonal mode, including fingernails and toenails. It can still render in Mental Ray if I do 1 level of subdivision, it can render in Renderman as is…
This is a Maya rendering, no raytracing, very fast. (See my recent foot modeling tute on my site.) It’s one piece, except for the head (which has to be interchangeable).


What does the wire look like smoothed?

Looks great but the arms are a bit boxy looking… if you know what I mean.


Originally posted by gmask
[B]What does the wire look like smoothed?

Looks great but the arms are a bit boxy looking… if you know what I mean. [/B]

thats what i was wondering

maybe just smooth it?

i can never get those sub’ds to do anything anyways…to me theyre a mess…


Impressive as usual. Funny you are going down the 3/5 sided route just as I have been converted to quads for edge-looping. Oh well, I figure as long as the model gets done and is easy to work on for whomever needs to, then you’re in good shape. Looks cool.


yinako, I know a lot of people agree with you, and I just don’t know why. I mean, I’ve never had any trouble with this method. See image below, this is the same polygonal cage converted into a SubD. I don’t think it looks very angular, and I think it will animate fine.


I guess this may not apply to subd, but when I smooth out a polygon base 1 or 2 times and do a triangulate to see how the surface renders out in maya, areas with 3 or 5 side will no longer follow the contour or topology you want(ie sometimes form series of diamonds.

All these are very minor details that usually can’t hurt when you view a render from far away, if you are doing localised details with 3 or 5 there isn’t mcsh worry. But to for main focused gemetry you would like to keep the tessellation predictable.


Fascinating. I’ve heard you say this before, Stahlberg, yet until these pictures it hadn’t really sunk in enough for me to experiment or change my workflow. I’m not OCD about killing all stray tri/ngons, but I would never leave some of the things you’ve left in your mesh, for example the loops that just terminate right into the side of a poly:

But I’ve never really questioned why? That’s just what I was told when I first began modeling and I accepted the edge-loop / quad approach as axiomatic, despite how enormously frustrating it can be.

Ok, I gotta apologize in advance for the following barrage of questions, I’m just very curious about your thinking here …

Do you think this is just a case of you mastering the rules to the point where you know when and where you can break them? Because you do have nice flowing edge-loops, it just seems that perhaps you know when and where you can get away with abruptly terminating detail in a way less experienced modelers would be more wary of.

And if this is the case have you made any useful distinctions on when/where you can get away with it?

For example, where on a face model would you feel comfortable with terminating detail abrupting, leaving tris/ngons or poles?

Can you think of a case where you make sure that you had all super clean edge-loops/quads? For example, maybe in an incredibly flexible character that needed to deform anywhere and possibly with skin/muscle sliding/bulging?

Also, would you suggest this approach to someone newer to modeling? Some may say it could lead to sloppy modeling habits, while on the other hand an overly technical focus on “pure” topology can be frustrating and drain the artistic process.

Finally … obviously Bay is an excellent modeler and possibly the original and most influential proponent of edge-loop / quad philosophy … so do you believe this is more of a personal workflow issue (that clearly works for him and many others) than a fundamental rule of modeling for proper shading/deformation? (I’m reminded of this thread:



You are my man, you use the words I have been searching for. I tried a few times to convince other CGTalkers that subdivisions ARE usable in Maya but I lack the experience of a good modeler.
At some point I thougt that my workflow was wrong but you just showed me that someone can get results modeling this way. (excellent model by the way).

For the interested, here is how I model things:

The beginning is very much the same as in stahlberg’s foot modeling tutorial. But I convert to Subdivisions right after I’ve defined my basic poly shape.

After conversion to subdivisions in polygon mode I put the poly cage and smoothed subdiv on separate layers (for easier vertex/edge selection on the poly cage or for uncluttered smoothed previewing of the model)

What I’ve found out that slows down subdivisions a lot is rerouting edges (don’t know if this is the correct word. In other words deleting an edge and afterwards resplitting the face, but the edge in another direction)

Don’t get me wrong I’m not trying to convert the quad face/edgeloop modelers to something (I think is) more fun and flowing way of modeling. (keeping all faces quad AND keep the model ultra light… I still haven’t figured it out)
I just wanted to let Stahlberg know that he’s not alone with his modeling thechnique.


when you say the model renders in PRman, what surfaces did you use? Maya subd’s or renderman sub’ds? I ask because I have had trouble when converting faces with more than 4 sides when using renderman…



Could you Mr. Stahlberg post wires from the back view?! thanks…


I’ve tryed to render many polygon models in Renderman as pixar subdivisions and it looks great, just like if you have smoothed your low cage many times but it’s smother of course. Because Maya Subdvs is not supported there.
I wonder why Mental Ray doesnt have such a tool in standard version!! it would be very cool to see such kind of surfaces in MR.
Eyad, you could see some wires of backs on Steven’s website
Thanks steven for saving our lifes !!


Great work Mr Stahlberg!! :thumbsup:

I have always liked Maya’s sub-d’s and have never understood why people say they are not any good. :shrug:

I think it’s great that you like/use 3/5 sided polys. I think it’s part of the advantage of using poly/sub-d’s over NURBS… but I know of at least one CG Studio that makes the modelers use only 4 sided polys because the studio still treats them like NURBS and the poly models need to be “patched out” for texturing.

Anyway, again GREAT work! :slight_smile:


Originally posted by Stahlberg
yinako, I know a lot of people agree with you, and I just don’t know why. I mean, I’ve never had any trouble with this method. See image below, this is the same polygonal cage converted into a SubD. I don’t think it looks very angular, and I think it will animate fine.

It’s mainly the inside of the elbow that looks square to me…

Could you post one smoothed version fo the wireframe? Would you mind also showing a polycage wireframe view of the back?

It may not make much of a difference but I think the edgeloop thing is more important for the face.

Also if the head is a seperate piece how are you keeping the seam from showing and still blendshapes? My approach would connect the head and use sets of vertices and blendshapes instead of shapes and blendshapes.



I have a couple of quick questions. You said in your ealier post that you body poly count was 3400 in polygonal mode. Do you mean 3400 faces unsmoothed or… ? I’m a bit confused since maya usually reports Verts/Edges/Faces/UV’s.

Also is that 3400 inlcuding the head, is that just from the neck down?

And finally, where do you seam the head/body together? Top of the neck or …?


Steven, with maya 5.01 u can now render low poly directly with mental Ray subd´s have u tried this? so if u have a good skinning u need no wrap deformer i have tried it and it looks good (texture smooths too)


swag, that sounds great. It gives me another rendering option.
JasonA, that’s 3400 faces unsmoothed, not including the head. The seam goes just under the jawbone from ear to ear.
gmask, I agree about the inside of the elbows. And about the seam, from what I’ve heard I might be able to weld the head on later and still keep all my Blendshapes for each surface, but that is un-tested. If not, I’m sure I can find some way to minimize the seam.
lazzhar, roger and TomD, thanks.
Elliotjnewman, I tested Maya SubD’s in MTOR.

somlor, I’m not sure if I’ve mastered anything. :slight_smile: I just keep going back and forth from the poly to the SubD, spinning it and comparing it to my references. If it doesn’t look right I try something different. Repeat until I’m either satisfied, or say ‘close enough for now’. (I’m never really done with these things you know.)
You could get away with leaving 5’s and 3’s in many places on the face, it all depends on the exact shape and what’s next to it. What creates a bump in the SubD or smoothed version is when a polygon has a big difference in it’s surface area compared to the ones next to it. Big/small, fat/thin, quad/tri - triangles are usually smaller than the neighbouring quads, that’s mostly what makes the bump or discontinuity. It’s all about how close 2 eges or 2 points are to each other, a lot like NURBS modeling. 5-sided can be more easily ‘hidden’ if you have to, they aren’t as likely to show up clearly as triangles. But they will if the surfaces deforms a lot, so watch out for that.

Yes, I can imagine some kind of perfectly smooth blob-like creature that couldn’t take any triangles or n-gons anywhere. But that’s theoretical.
I’d teach this method to a beginner, because it’s easier than forcing all-quads. You can always struggle with that all-quad requirement later if you had to - although I would introduce the methods, how to do it all-quads, too (you could call my method ‘mostly quads’). I’m not against edgeloops per se,they’re a vital concept as fas as placing the edges to follow surface contours and wrinkles etc. But they don’t necessarily IMO have to form perfect loops everytime (in Maya).
I’m not that familiar with Bay’s modeling but I believe he started modeling on a software that required all-quads right (Mirai)? Anyway, if I remember correctly I think I’ve seen some example of his that wasn’t totally all-quad. But I’m not sure. But, yes it obviously works very well for him. So it could be more of a personal choice I suppose.
eYadNesS, here’s the back wire.


What I understood so far about edgeloop is that you have to try to avoid pole, but tris and quad could coexist if there is no troubles when animating/rendering.
Maybe this thread found in spiraloid forum could help!
Anyway very very interesting topic:)


yeah,great stuff
the topology is really following/flowing along the basic muscle/bone structures…
and really even the form of the body like the “buns”(lol)…etc
intresting stuff steve,thx


Steven can you post this on spiraloid i think it would make a very interesting discussion.



to swag
can you please shed some more light on that direct rendering
of low poly surface with mental ray?
Does it mean that I dont have to convert my lowpoly in to subdivision surface
and that I can tell mray to subdivide it at render time?
That one would be a life saver.