>>2844601You can disable physical calculation for individual bones during the motion, though it might not work super well when done on the go. The processing of physics also differs between MMD and MMM. I recall seeing a Japanese MMDer on Twitter say that you need to first register a key with physics on, then with physics turned off, and vice versa for turning them back on. MMM allows you to turn them on an off on the fly during playback, but After Physics weights freeze in place and won't move without physics, and baked keys for controlled behavior won't work without disabling physics for the model—they'll just stay frozen in the position where the physics got turned off until you turn them back on. The skirt should be already invisible in this case though.
>>2844528Wow, it actually works now, although it starts to deviate from the supposed target (grip) when the arm is raised (pic related), but that can be partially mitigated by matching new IK bone positions to the intended real bone positions (rotated by 36 degrees). Thank you.
The rotation offset does work for transferring the IK state to regular arm bones. Are small XY angle adjustments important though? It's one thing to apply -36 on Z but another entirely to have to copy three precise values. I know that zero rotation on an axis can lead to snapping.