Create a rough pass as fast as possible to get the entire scope of your work functioning.

Automate as much as you can so changes / iterations are fast and easy.

Now things are going! ;)

This setup step looks as simple as it is - but it renders! Now we can start placing the 3D geometry to suit.

First, I place the eyes. This is just the setup, so I'm using a simple (NURBS) sphere for each eye (referenced, of course) that uses an aim-at constraint on a locator that I can place behind the camera.

Because I have a number of human character tools around, I create an entire human character skeletal structure with NURBS 'blobs' parented to the skeleton for reference as an initial starting point. Basically, this looks like this:

• Character File
  • Reference to skeleton file (with full human hierarchy - backbones, hands, etc.)
  • Reference to skin geometry (turned off)
  • Reference to head geometry (turned off)
  • Reference to eye geometry (placeholder NURBS selected and parented to skeleton)
  • Reference to hair setup (turned off)
  • Reference to clothing

This setup isn't as complex as it looks. If done initially, it's easy to setup AND provides fast flexibility to make HUGE changes later when the unknown is kicking my ass. ;) So, I'm going with it.

An example of what this really means is creating a low level element (like an eye) and saving it. Then, you go into your character scene file and create a locator (say, LeyePos), then create a reference (read in the eye file) and make it a child of the LeyePos. Now you can move the Pos (position) locator in space - when you want to change the eye geometry, you go back to the eye file - save changes as a different file. When you reload your character file, you use the reference editor to 'replace reference' with the new eye geometry file - bam! Your character has the new eye geometry - and it's already in the right place / hierarchy, etc.

Later, if you have a problem rendering - you know the bug is in your new eye file and it'll be easy to track down / fix.

In the above image, I've just used joint rotations to assume a pose that the subject possibly was in, using the blob model (spheres parented to the joints). The 'set' geometry is just placeholder as well, but it has my default grid texture to give initial UV feedback.

The important part: The joint rotations are in the scene file. When I open a new file and reference in the character (blob skeleton) again - the skeleton itself has no key rotations. For the purposes of creating skin geometry, I pose the skeleton in a position that makes creating skin geometry the easiest and then export out ONLY the skin. When I use (reference) the skin geometry in my character file - I can set it up with the most control. Returning to the scene file - the new skin is used in the final pose automatically. More on this later.

With a little tweaking - I position the camera (using 205mm lens - a guess) and tune the pose of the character until I get a close match - as so:

Ack, Ack - Anybody?

In the view above, I have most of the reference image set to be transparent - you can see the light gray where the HD aspect ratio is visible.

In this next view, what I'm doing becomes clearer. The 3D skeleton is positioned to match the original pose and I end up getting exact information about where the skull was and how it was placed with the neck, shoulders, arms, jaw and even cheekbones lining up.

This view is less obvious, but still useful. Not just confirmation of the quality of setup, but the shading contours match the arm, chest, shoulder, neck and back. The background angle still looks a little off, so I'll reposition the background geometry. Almost ready to model.

And yes, it renders!