Retarget kinefx non biped to biped- issue

   735   3   2
User Avatar
Member
4 posts
Joined: April 2022
Offline
Hello everyone,
I’m having some difficulties properly transferring old humanoid animations (+150) that are non-biped onto a biped mesh/skeleton (coming from CC4).
The retargeting itself works fine (after isolating the correct bones), but depending on the imported clips, I still end up with some issues where a joint "jumps" on certain frames of the clip. These only occur on the target skeleton (after retargeting).

I’ve managed to fix some of them via rig pose, but that’s not great and far from optimal given the number of animations to transfer. It’s probably a Gimbal/Euler issue, but since I’ve never really touched that part in Houdini, I’m a bit lost in front of all the possibilities the software offers to modify them!

Would you have a clear workflow on how to fix these rotation issues or joint/bone jumps on certain frames post-retarget (or even before)?

Thanks everyone.
User Avatar
Member
8155 posts
Joined: July 2005
Offline
It's really hard to say without actually looking at exactly the case when it happens. There a number of potential causes:
- The rotation jumps come from the mocap/"old humanoid animations" themselves. I've seen my share of these from mocap.
- If the mocap/animation you're retargeting to moves to an 180 direction from the T pose (used when doing the matching), then this can cause issues. In this case, use the Root Constrain feature on the Full Body IK node.
- It might be there are insufficient iterations on the Full Body IK node. You can try putting this to a large number to see if fixes the particular case, in conjunction with using the Root Constrain.
User Avatar
Member
4 posts
Joined: April 2022
Offline
Hello Edward,

Thank you for your reply. Indeed, the “anomalies” on the clips are quite erratic.
From what I understand, if I cancel rotations on certain bone axes at specific keyframes, The jitters disappear.
The problem is that these jitters only occur after retargeting.
And on the FBIK solver side, I’m already at 50 to 100 iterations — so I have to check each clip both before and after retargeting, which is extremely slow and time-consuming.
Thanks for the info about the root constraint node.
I need to look into its use.
User Avatar
Member
8155 posts
Joined: July 2005
Offline
TDy
From what I understand, if I cancel rotations on certain bone axes at specific keyframes, The jitters disappear.

Maybe you could do that procedurally? If you know certain bones should never have certain rotations, then you can probably do it with some combination of attrib wrangling in conjunction with Attribute Transform Extract and Attribute Transform Combine SOPs.
  • Quick Links