Modelling for Animation

Zephir62

I've always had trouble with figuring out the best methods in modelling for animation. PoopinmyMouth's little tutorial was a little confusing:

http://www.poopinmymouth.com/process/tips/limb_deformations.gif

Wouldn't the first example produce an awkward animation with the leg visibly cutting into itself?

If you notice in this video:

http://download.autodesk.com/media/MNE/ms/skin_mirror_150k.wmv

The leg bends perfectly and it doesn't cut into itself as shown in Poop's tutorial. How is this done?

Also, what are methods for bending in the back/arms/hands/etc? I'm sure the answer would be too long for each of them, but there has to be some sort of general rules to follow or help guide in the process.. Any help?

Rick Stirling

Poops is more or less right

Because you are dealing with polygons, and not deforming muscle mass, you basically have 2 options - pipecleaner mass loss, or intersection. Intersection works best in most cases.

CheeseOnToast

Take a look at the .pdf swordmaster tutorials from http://www.xcloud.net/. The knee and elbow construction works really nicely. It's in Korean I think, but the pictures are enough.

Noren

Poops example is correct, but a little simplified.
The deformation also highly depends on placement of the joint in the limb and the positions of the vertices like the xcloud example shows. Even a one loop solution can look ok to a certain angle for meshes with very low polycount if you model the loop in a 45 degree angle like if the joint was already bent 90 degree. (If you know what I mean. )
Ideally you have driven extra bones that you can assign those "middle" loops to, so there will be no volume loss.If you don't have that possibility and you have to deal with a standard rig, you should in most cases also try to keep your mesh simple. That goes hand in hand most of the time anyway.
Most people tend to think that more polygons/loops mean better deformation but often the opposite is the case.
Intersections look smoother of course with higher polycounts.
I tend to quickly skin my chars before I start the uv-maps, in order to test the deformations. You can predict those to a certain degree, especially with simple loop based geometry and by experience. But expecially if you get near the axis of the joint, with different weightings etc., things get a bit unpredictable, at least for me.