We've been looking into character animation in 3D Studio MAX over the last few lessons, devoting our attention to setting up a viable skeleton and skinning it with a low-poly character mesh. It's been slow, tedious work, but we did it.

Of course, the process began with modeling the character mesh, but we jumped over this subject. I want to return to it now for a couple of lessons. We'll be getting into some deep and general material here that should be of interest to those who use packages other than MAX.

It goes without saying that, whatever its animation powers, 3D Studio MAX cannot compete as a character animation platform if its character modeling toolset is unsatisfactory. MAX's history as a modeling platform is complex, but a sense of it is essential to the serious user. Let's work through some background that should prepare you for the modeling ideas I'll be offering.

The original version of MAX was closely tied to a unique vision of modeling based on parametric primitive objects (boxes, spheres, etc.) subject to a modifier stack. Whatever value this system had for modeling simple manufactured items, it was completely inadequate to the demands of organic and character modeling.

Organic character modeling was proceeding down two parallel paths in competing programs. The high-end packages, Softimage and Alias PowerAnimator (the ancestor of today's Maya) were the first to offer NURBS modeling, and most users of these programs were convinced that NURBS was the future of character work. On the other hand, Lightwave perfected a method of smoothing polygonal models that was functionally similar to NURBS – so much so that they called it MetaNURBS.

Smoothed polygonal modeling, as implemented in MetaNURBS, has defied the expectations of many that NURBS modeling would ultimately triumph in the character world. Today, that world is profoundly split between both approaches. With an increase in its stature, smoothed polygonal modeling acquired a new name – subdivision surfaces. Maya – a bellwether application in all respects – treats subdivision surfaces as the co-equal of NURBS in its most recent version.

For the most part, I'm with those who prefer subdivision surfaces to NURBS for character modeling. The reason is very simple. Branching architecture is difficult and problematic with NURBS surfaces. To create, for example, a satisfactory arm branching out a torso, you must preserve continuity between the NURBS surfaces that meet at the shoulder. Blend-type surfaces are less than ideal here. True NURBS surface continuity between patches is possible in programs like Maya and Softimage, but it takes work and a lot of experience to get it right. Don't misunderstand. I'm a strong believer in NURBS modeling and in MAX NURBS in particular, as must be evident from the columns I've written about the subject in the past. But I'm less than enthusiastic about NURBS for general character modeling.

I don't want to get off on a detailed explanation of the problems with NURBS modeling for branching geometry. It's enough to say that the problem doesn't exist at all with subdivision surfaces, as we'll see. Lightwave has strong subdivision surface tools, and Maya's are very impressive as well. Softimage completely missed the boat on this subject, contributing to the general sense that this package is no longer remaining competitive. But what about MAX?