The Absolute Basics of Rigging
If you're alive and want to move, you're going to need a skeleton. If you're working in 3D and want to move 3D things, they're going to need the digital equivalent - a rig.
A set of connected bones is just half of a rig though - the other half is having controllers to move it. Think of it like a traditional rod puppet; yes it needs to be structurally stable, but you also need something to move it.
So...one half bones, one half controllers, but the final half is making sure that it looks good when you animate it. That's largely do to "weighting", which is how much each bone controls each area.
Read on, and learn all three halves and a few more quarters
Rig bones aren't too different to human bones - if you have the right ones in the right place, they'll move the bones. Get one in the wrong place, and it's going to give you a headache.
Having a good understanding of anatomy will help (I will give my artists VERY precise instructions as to the placement of bones), but as long as you get them roughly right, you'll probably be fine
If you want to move a person, do you peel off their skin and grab their bones? Most of you wouldn't. It's rude, but it's also difficult to do, bones are generally fairly well hidden.
It's the same in 3D. Grabbing bones is awkward, so we make controllers to make it easier. Having controllers is useful for a bunch more reasons though - they act as a buffer between the bones and the animator, so that if the bones need to change for some reason, they can change without the animation being lost (since the animation is all on the controllers, not the bones). They can also carry attributes - so a hand controller might have a "fist" attribute, which poses the hand into a fist, and saves the animator the time of posing each joint of each finger manually
Animating on raw joints
- Tiny click targets, buried inside the mesh
- Easy to grab the wrong one when several stack close together
- Nothing stops an animator from dragging a joint out of its bind pose
Animating on controllers
- Big, clear, color-coded shapes - easy to grab even zoomed out
- Keyframes live on the control, not the joint - a rigger can fix the skeleton later without touching the animation
- The control sits at the joint it rotates, and can limit or relabel exactly what an animator is allowed to touch
Forward Kinematics vs Inverse Kinematics. Sounds like horrible gross tech stuff, but it's pretty important in rigging. Forward kinematics is moving one bone at a time, like a stop motion puppet. Inverse Kinematics is grabbing the end of a chain of bones and moving that, like a Sesame Street style rod puppet.
Both are useful in different scenarios, so most rigs will use both, and roll in an FK/IK switch, so you can jump or even blend between them
Forward Kinematics
Rotations flow parent to child. The hand ends up wherever the chain of rotations puts it.
- Move the shoulder and the hand drags off the wall - no constraint holds it
- Great for arcs, tails, capes - anything with a free end
Inverse Kinematics
The hand is pinned to a goal. The chain solves backwards to keep it there.
- Move the shoulder and the arm re-solves around the planted hand
- Drag the diamond (called a pole vector) to choose which way the elbow bends
A tentacle, a wire, a tail - anything that can wrap around something is still going to need joints, but it'll need a lot of them. We call these joint chains. Again, IK vs FK is a big decision, and the correct answer depends on the use case scenario (general flapping is where IK wins, specific grabbing is FK territory), although often you won't have one control per joint, because too many controllers makes animators angry
FK on a chain
Every control is a rotation. Spans inherit their control's rotation, so one ring curls several bones.
- The tip never chases your pointer - it lands wherever the rotations put it
- Fewer controls, same smooth curl: that's the follow/inherit trick real tentacle rigs use
IK on a chain
No rotating at all - drag the tip and the whole chain solves to reach it.
- Good for reaching, whipping, or tracking a moving object
- Feel the difference: IK pins the end, FK never does
Weight painting is a weird one, since it doesn't really have any real life parallel. You're basically telling the software which part of the model should be controlled by which bone. Every vertex on the model is "weighted" towards one or more bones. So hopefully the vertices on the skull are 100% weighted to the skull bone. But if 1% of one of them is weighted to a foot, then every time the character moves his foot, part of his skull will move.
Where it becomes tricky is around the joints, since the skin there in real life would be affected by multiple bones. Often what happens is that people just give those vertices equal weighting of all bones in that area, but that often results in soft and rubbery boneless feeling joints.
You'll never notice it if it's done well, but if it's done badly, things get pretty broken pretty quickly.
Too little blend
A hard 0/1 split right at the joint, nothing smooths the transition between bones.
- The mesh visibly splits open as the joint bends
- Try "Hard edge," then bend with the controller
Too much blend
Every point gets pulled by both bones at once, everywhere along the strip.
- The joint collapses inward - the classic "candy wrapper" look
- Try "Over-blend," then bend with the controller