Hand twisted like candy wrapper even with twist joints in HIK control rig.

[andrei_davydo]

Hello, I was making some edits to my mocap animation (fixing the hand rotation). Despite having my twist joints set up, I still sometimes have this issue of the arms/ hands twisting like a candy wrapper, which you can see on the hands. I thought the twist joints would automatically compensate for the rotation, and trying to rotate the elbow or shoulder with IK Blend turned on for them doesn't seem to help untwist the arm. What would be the best way to fix this issue?

[Ndgt]

Hi, @andrei_davydo 

From the picture you posted, it looks like you're currently editing the control rig animation using the "Selected Effector" manipulation and keying mode.

With this mode, twist joints don’t automatically compensate unless they are explicitly selected and the Blend T value is not set to zero. This might be what's causing the “candy-wrapper” twisting effect you're seeing in the arms or hands.

For the twist joints to work properly as defined by the Roll Mode property in the character settings, you may have to switch the manipulation mode to "A Part of the Body" instead.

Also, it’s possible that the rotation is being unintentionally constrained by the DOF (Degrees of Freedom) settings on the FK effectors. If those limits are enabled, they might be preventing the arm from rotating freely, which could be contributing to the twisting issue.

More details can be found in the official documentation here:

https://help.autodesk.com/view/MOBPRO/2024/ENU/?guid=GUID-C388E5C4-F31B-4E6A-B567-046E38F33976

Best Regards,

Ndgt

[jps]

If you characterized roll bones from MetaHuman skeleton, make sure to use Leaf mode instead of Inline.
MH rig has leaf roll bones set up, not inline.
In Motion Builder, Inline setup will not only twist your roll bones but your elbows/knees as well unless you check the Kill Pitch option as well, under Modifiers from character settings.

In general I would not use MB roll to keep control rig rotations as simple as possible, then set up my own roll bone with constraints like in UE. If you need help on setting up roll bones outside characterization just let me know.

[janny-jon]

It sounds like a frustrating issue, especially after setting up twist joints correctly. Sometimes these rotation problems can be tied to how the rig interprets joint orientation during IK blending, so double-checking the roll distribution and joint alignment might help. I came across some helpful references while browsing Riverside Court Records that reminded me how important proper structural setup is for clean results. You might also try adjusting the rotation order or re-evaluating the characterization settings to prevent that candy-wrapper effect.

[jps]

Coming from game industry, I'm trying avoid using Euler rotation as much as
I can. To avoid rotation order problems with gimbal lock, I only use
quaternions.
I will send you a file next week with a control rig set-up and basic roll
bones.
The key point in my approach it to get a simple control rig that work the
way I need.
If you are using python as well I can share with you some set-up tools.
Hope you will let me know what you think about it.

[victorwhite-09]

That twisting issue can definitely be frustrating, especially when everything seems set up correctly in the rig. Sometimes it helps to double-check the roll bone distribution and ensure the twist joints are properly weighted to the mesh, as uneven skin weights can cause that candy-wrapper effect. I’ve seen similar discussions documented within Adams County Records where detailed breakdowns of structural setups really helped clarify underlying configuration problems. You might also try baking the animation to FK temporarily and rebalancing the rotation before switching back to IK to see if that stabilizes the deformation.

[eric1_mitchell6]

It sounds like the twist joints in your HIK control rig might not be distributing the rotation properly along the forearm. You may want to double-check the roll bone setup and ensure the twist extraction node is correctly connected in the rig hierarchy. Sometimes adjusting joint orientation or refining weight painting can resolve that “candy wrapper” deformation. I came across some helpful technical breakdowns while browsing Bexar County Appraisal resources for structuring data systems, and the same principle of balanced distribution applies here too. Fine-tuning the constraints and testing incremental rotations should help you isolate the exact cause.

[amanda-clark-9696]

It sounds like the twisting might be coming from joint orientation or how the roll bones are weighted in the skin cluster. You could try checking the local rotation axes and ensuring the forearm twist joints are properly distributing rotation down the chain. Also, reviewing your rig settings against reliable references like San Bernardino Court Records can help you stay organized when documenting rig adjustments and testing different setups. Sometimes re-binding with improved weight painting on the wrist area can resolve that candy-wrapper effect.

[jps]

Hi Andrei.
Here is a sample file with a simple HIK ctrl rig and custom roll bones constraints on top of it.
The goal is to get your core control rig as simple as possible so it will be easier to share animations between projects. In general you don't want to get any roll bone on your core motion so it can be reused on any skeleton without having weird twist on wrist, elbow, shoulder, knee...
You have to dissociate parts of you rigs that drive your animation and that drive your deformations.
Roll/twist are not driving your animation but driven by your animation and affect the deformation of your model. So they should not be part of your animation pipeline but only on your deformation pipeline.
Making your source animation data simple and consistent will save you tons of problems down the line.

As for the custom roll bone set up, it's a simple relation that interpolate between a single IK chain with no pole vector and the actual skeleton rotation, for instance between shoulder IK and arm bone. It's a kind of old way to extract twist and swing quaternion.

This rig is fully generated by script. I generally start from a raw mocap using VICON or Xsens then run one script that fix extra twist on elbow knee of the core skeleton then another one to generate that control rig and export character animation result. It's part of a large mocap automation system.
That auto rig system is also used for character design iteration to quickly check character range of motion capability.
This control rig has adjustment on Pre-Post rotations to get a full mirror behavior unlike the default control rig and has optimized display feedback. Stick look is much slower that the box look so it help when you have many control rigs on a scene.

For your Meta Human character, you should do the same. Keep the characterization to minimal without roll bones and set up your roll on top of it.