Another Dummy, This Time With Blender

A couple weeks ago I put up a post about creating a simple, rigged dummy in POV-ray. Though it worked more or less I was ultimately unsatisfied with the result. The process of posing the figure was error-prone and time-consuming, and ultimately it just didn't look quite right. So I decided to revisit the idea of building a dummy using Blender since it's designed for just this sort of exercise.

Blender's a great tool, but it has to have the steepest learning curve of just about any program I've ever used. The 2.5 redesign is a vast improvement over v2.4, making the program merely difficult where it used to be absolutely inscrutable. The downside is that the 2.5 documentation is still woefully incomplete, making it a challenge to use some of the more advanced features. So now, a couple of weeks later, and after much reading and experimentation, I've been able to put together a rigged dummy:

It took a lot longer than in POV-ray due, in large part, to my general unfamiliarity with Blender. But I also took a lot of time modeling the hands, something I didn't even attempt with POV-ray, because I'm going to need reasonably realistic hand geometry for some of the things which I intend to do with the dummy. As far as actually posing the dummy Blender is light-years ahead of POV-ray. Posing naturally constrains limb lengths (a process which I had to fake in POV-ray using vtransform and Rotate_Around_Trans) thus producing a much better-looking result. And you get immediate, OpenGL-rendered feedback rather than having to go through a render cycle. Blender is truly more suited for model-and-pose work than POV-ray, provided that you're willing to put in the (considerable) time to learn how to use it.

If you'll recall from my previous post I started off with POV-ray because I wanted to be able to quickly render multiple camera angles from the command line. Blender, it turns out, can do the same thing using its ability to store multiple "scenes" in a single file. Here's what I did to get the four-shot composite above:

1. Create a Blender file containing four scenes, one for each of the four shots that I want to produce.
2. Give each scene a unique camera and light source positioned as desired.
3. Create one dummy model and one floor grid (grid texture provided by an image, btw) and link them to all four scenes.
4. Render each scene on the command line as follows: blender -b <Blender file> -S "<scene name>" -o <output file name> -F <format> -x 1 -f 1
5. Composite using montage.