Artsy Rants

3D Character Modeling: Working in Virtual Clay

  • March 3, 2015

I’m frequently asked: exactly how does one “draw” a 3D model? The first time I was brought up short as I tried to explain it with words and hand-waving; it’s not something I’d thought of in that way, it’s just something I do.

Now that I’ve fielded the question a few times I have a sort of “elevator pitch” developed after various puzzled looks told me I wasn’t getting the idea across. Now I draw an analogy to “real” sculpting, and the medium I try to leverage is modeling in clay.

When a sculptor sits down to begin a project, often as not they start by plopping down a formless blob of clay. It might be a block from a package, an amorphous handful they’ve pulled from a larger source, or something else entirely. The common factor is it probably starts out looking nothing at all like the finished product.

So goes it with the style of modeling I favor, sometimes called “box modeling” because that’s a very common shape to begin with. Actually, I tend to cheat a little, because in most cases the software you’re modeling in will give you a rather nice selection of primitives to use as your starting point, and odds are you’ll be able to find one that (very roughly) matches where you’re going.

Take my usual starting primitive for a humanoid torso, for example: a cylinder. From experience I already know how many sides and segments I’m going to want for starters, and so my “blob of clay” starts out looking something like this:

polygon primitive

And now the fun begins! Much like that metaphoric sculptor with their clay, I start by pulling bits of the shape around to be closer to the form I want. In 3D modeling those “bits” are the points that define the shape, and the polygons connecting those points. With most 3D modeling tools you can interactively “grab and drag” portions of the shape, switching from view to view and rotating your viewpoint as needed to see what effect you’re having.

Of course the primitive isn’t going to have enough geometry (collective term for all the points and polygons) to form the final shapes, so you use other 3D modeling tools to cut, add, delete and otherwise fold, bend, spindle and mutilate the existing polygon structure as you go. As the shape gets more and more complicated, so does the task of keeping the “polygon flow” tidy and apportioned well. One primary consideration when modeling a character is that it will almost certainly be animated, and there’s a black art to making sure various areas on the model have polygons in arrangement and number to facilitate clean bending and other deformation.

Here is a basic torso shape created from the primitive shown above:

polygon cage

Now, there’s another consideration that is often important in character modeling: something called (in Lightwave, other software can differ) “subdivision surfaces.” As you can see, the torso is pretty blocky and jaggy looking; not very appealing. We’re going to fix that, but first we have to make sure the entire object we’re building is composed of only “quads” and “tris,” that is, polygons with exactly three or four points. For technical reasons tied to the underlying math of how subdivision surfaces are calculated “quads” are preferred but sometimes you just have to slip a “tri” in here and there, especially when increasing or decreasing the number of polys (polygons) around the circumference of a rounded shape.

Assuming you’ve met that restriction, going “subD” with your model is a matter simply of turning on the display mode. The software semi-magically subdivides that blocky poly cage and yields something much more pleasing to the eye:

subdivided model

Bear in mind the underlying “poly cage” has not been changed one bit: you can switch back and forth between the cage and the subD version as easily as tapping a key. Many experienced character modelers will quickly switch to subD mode and do most or all of the shaping directly on the smoothed version rather than manipulating the cage and flipping back and forth.

I personally prefer working with the subD display once the basic form is blocked out, but I also flip back now and then to see whether the points of the underlying cage are getting pulled into weird contortions. It’s just a personal quirk of mine that dictates the cage should be somewhat tidy if possible. When you are sculpting in subD mode it’s common for the underlying points to get pulled any which-a-way. I like to try to “un-mangle” my cages if it’s possible while still retaining the desired subD shape. Many fabulous modelers would tell me (I expect, rightly) my quirk is a waste of effort. What can I say? Artistic license.

And that’s how it works. In this case my next step would be to drop another primitive to start forming a leg and, once that’s done, mirror (make a mirror image copy for the other leg) and attach them. Then the arms, and so on.

Hope this has been interesting, maybe even a little bit fun to read. Please leave comments below, and share this post far and wide. You can see examples of my work on my main web site, Go take a look, I’d like that.


Spinland Studios, LLC is a high-tech 3D modeling and animation studio in the Mohawk Valley of Upstate New York. Defy conventional branding and let me bring your brand to life! Visit for more information and examples.

Spinland Studios, LLC is a proud member of these organizations (click each to learn more)