hmm so the solution is to have another “duplicate” armature of this “retarget rig (mixamo)”.
And then retarget again from “retarget rig (mixamo)” bones to this new “duplicated” armature bones.
Hmm I think it won’t work as the “animation data (keyframes)” is residing on “Capture Armature” that I can “bake action” on. And I can’t remove the contraints on “retarget Rig (Mixamo)” else it will break the “follow motion” and I still can’t bake “action” on this “retarget Rig (Mixamo)”.
Also I need to “learn” retargeting animation data from one rig to another … which might be more like .
retarget “Capture Armature” (that has the animation data) to “retarget Rig (Mixamo)” … that could be what I think you are saying.
Hi bandages,
Still using Blender 2.92 … with “keemap retargeting” addon.
I have been trying to “retarget” this “capture rig” (that has the animation data) to another mixamo rig - although the animation get copied over, but the mesh attached to the “retarget rig” (mixamo rig) get twisted up.
My approach based on what I read, that I should match the source rig and the retarget rig with similar pose position, and also the bones rotation matches. When ever I get these rotation of bones matches (in this cases both set of shoulders, arms and hands, the mesh that attached to the retarget rig (Mixamo) get all twisted up.
Am I doing right way - that is to rotate the bones of my retarget rig in pose mode to match with the source rig before I should copy the animation data? Or should it be edited mode (which I find even more messy on twisted mesh of the retarget rig).
Please help as I am all confused now about this rotation of bones, how exactly should I be doing it.
See attached my latest file with my best attempt but still both hands of retarget rigs are twisted.
Been trying this for a few days without success … except the animation gets copied over but with twisted hands !
Duplicate the armature with all the constraints. Bake with clear constraints. Switch to dope sheet/action editor, name the action, give it a fake user. Delete the rig. Capture another animation. Repeat. Eventually, you have a library of actions. All of these actions are usable on any copy of the rig without constraints.
Your animation data-- your markers, the points that your bones are stretching to-- don’t contain information about twist. They contain information about where your wrist is, but not which way the bottom of your wrist faces. Yes, reposing the rig in pose mode to point more accurately at your “rest pose” markers is fine, and will minimize rotation, but if you have ugly twist, you have to fix that manually by rotating the bones on the constrained rig, individually, in their local Y axes, and keyframing those values-- you have to do the animation there, because your capture doesn’t have that data.
Hi Bandages,
Thanks for responding so promptly.
I finally manage to do this part … appears I need to “tick” (enable) visual keying as well to make this works. My setup was disabled … can’t remember if it was default or I disabled along the way of testing
I repeated this for another animation and able to bake an action with this “mixamo” rig as well.
Thanks.
Well I tried to also repeat this with MMD rig that has IK bones. I had disabled all the IK bones.
It works for the top part of the body from torso (hip) upwards as I notice the x, y, z rotation matches with the “capture rig”.
However the lower part of the body, thigh, legs and foot … the z axis is opposite between this MMD rig and the capture rig. How do I change the MMD rig to match it to the capture rig for this z-axis?
I had tried in both pose and edit mode, but my mesh get twisted up.
See attached … I am still confused how to “rotate bones” on the re-target rig (in this case MMD rig with IK bones) without twisting up the mesh. Hope you can advise …
Attaching this MMD rig … in my best attempt … with twisted legs.
At rest pose - z in opposite direction only for legs (body rotation matches)
Hi Bandages,
Could advise me how you “attached” the monkey (Suzanne) to the bones ?
Is it by parenting Suzanne to the bone? I think you are right that I need the monkey to help me see better the rotation axis.
I really am in a lost after trying to visualize where this rotation axis is going wrong between the “capture rig” and the “retarget rig (MMD in this case now)”
ok found the problem.
Appears the source “capture rig” bone - both left/right thigh needed to add in the lock track as you have done on the “torso”. I was focusing on the “retarget rig” which was why I could not fix this problem no matter how and which constraints I’d applied
Thanks.
