Skinned animation usually is key-framed...

Do you mean vertex-morph/blend-shape animation VS skinned animation?

Sorry, I could've been more descriptive in my first post...

I'm confused now. I thought that skin-and-bone animation was where you define some skeleton with joints and then have some skin over it. As the skeleton moves you can calculate the movement needed for each vertex of the skin that is associated with that joint.

For keyframe animation, can't the artist define a bunch of different frames for certain times and then you linearly interpolate between the frames depending on your current time.

Those are both correct -- now, put them together and you've got key-framed skinned animation ;)

i.e. you define a skeleton with joints. The artist defines a bunch of different frames (which is a pose for the skeleton at a specific time) and you can linearly interpolate between those frames to generate in-between poses depending on the current time.
Once you've got a pose, you can use that to calculate the movement needed for each vertex of the skin (depending on the vertexes associated joint(s)).

This is the most common way of animating characters in games.

Start with skin-and-bone animation, but plan ahead so you can support full or partial keyframes for specific poses as well.

The big pluses of skeletal animations are IMHO:

• It is much easier to make animated models
• You need less data
• You can easily add IK solvers and adjust poses dynamically

Keyframes can give you better quality - but this implies you can invest enough into creating models and animations.

The big minus of skeletal animation is that a lot of the time it needs to run on the CPU. You can do keyframed animation (including interpolation) entirely on the GPU and with very high performance, although obviously there is the storage cost to consider.

The big minus of skeletal animation is that a lot of the time it needs to run on the CPU. You can do keyframed animation (including interpolation) entirely on the GPU and with very high performance, although obviously there is the storage cost to consider.

Skeletal animation can be done entierly on the GPU using vertex shaders (Look up hardware skinning, its extremely straightforward and the shaders aren't very expensive), transformation matrices for the bones can be stored on the GPU (which means you can store keyframes for the skeleton on the GPU aswell.

Some types of animations are excessivly difficult to make using skeletons though so there are definitly good reasons to use vertex morphing instead in some situations.

Skeletal animation works best for objects that have skeletons or joints, while vertex morphing works better for things that change shape (Turning a cylinder into a cube using skeletal animation is painful to say the least and forcing your artists to make such an animation using skeletons and vertex weights will drive them insane, Making a humanoid walk properly however is far easier to do using a skeleton and if you want something like ragdolls its pretty much required)

Mixing multiple animation techniques for a single object is possible aswell. (You can morph the vertices you use as the skin for a skeleton to create some pretty cool things)