The shoulder is the part of a biped rig that humbles people. Everyone can wire a clean two-bone IK leg. Far fewer can make a clavicle rotate convincingly as the arm lifts, automatically, without the animator babysitting it. This early prototype goes after exactly that problem.
The FK/IK limb logic first
Before the clavicle, the arm needs a limb solution that blends cleanly between FK and IK. That means duplicate chains, a blend driver, and a pole vector that doesn't flip. Get this wrong and the auto-clavicle has nothing stable to react to. The breakdown builds this foundational logic before touching the shoulder.
Cracking the auto-clavicle
The clavicle is hard because it should respond to the arm without taking control away from the animator. The approach: drive a base rotation from the arm's position so the shoulder lifts naturally, then leave an additive layer on top so the animator can push or dial it back. Automatic, but never locked.
A prototype, honestly
This is an early build, and it reads like one. The point of sharing it isn't a finished product; it's the reasoning, why the shoulder complex resists automation, and what a workable first solution looks like. The current framework's shoulder setup grew directly out of this experiment.