Retargeting seems backwards. Why do we need to duplicate animations for each different skeleton. Why can’t we modify the skeleton to use the original animations? It works like that in the IK Retargeter. Am I missing something?
the engine ties the Animation to the Skeletal rig, and the Skeletal rig (what you are probably referring to as “skeleton”) is the defining characteristic of a Skeletal mesh even if it doesn’t have any bones in the rig.
think of the animation as a set of deformation instructions for the vertices of the mesh usually expressed as change in position by virtue of changing the rotation of the bone. so that is why the animation is bound to the Skeletal Rig because if Ring_Finger_R_1 doesn’t exist then it would be very difficult to move it by <.2, -.1, 0>
the retargeting process attempts to map on Skeletal Rig onto another by creating the retargeter itself, then the engine uses the Retargeter to convert the given animation to the second Skeletal Rig,
effectively the steps the Engine takes:
- Anim01 is broken down into the deformations of the bones
- the deformations of the bones are re-evaluated by the difference in the default poses as dictated by the Retargeter
- the deformations are then reinterpreted onto the target Skeletal Rig
- a new Animation is created for the target Skeletal Rig
if you wanted to modify the Skeleton to use the original animation then that is supposed to be done in a 3D modeling/Animation program not in Unreal. the Engine cannot create, or directly modify Skeletal Rigs at this time, but it can attempt to interpolate from one Skeletal Rig to another through a retargeter.
the difference for IK Retargeter is that IK Retargeter only really cares about the Skeletal Rig itself, and the Goals/Constraints of the Skeletal Rig, and from that data it creates a micro animation in the form of a blend.