tamte
to see the usage you can look inside of Parent Constraint VOP
also it's docs [www.sidefx.com] describe "parentbind" better: "The bind pose of the current parent."
and also the Update Offset and Clear offset help describes what's going on under the hood, so essentially parentbind and newparentbind are the transforms at which the constrained object is "bound" to them

matrix offset = Value1 * invert(Value3);
outparent = offset * Value2;
outparent = blendtransforms(Value, outparent, 1, 2);

kodra
I kinda get it now. I think in 99% of the cases one should just use Parent Constraint SOP. Parent Blend SOP is more like an internal utility node

kodra
It's still a bit weird to me that how the algorithm works tho... cause the "blend" happens twice (one on parent, the other on the final result).

tamte
- it's very ugly to have stashed transforms inside of VOP network and also forced editable nodes

kodra

tamte
- it's very ugly to have stashed transforms inside of VOP network and also forced editable nodes

I guess this is one of the cases only a system like APEX can solve. (Blender's equivalent has the exact same problem: it needs you to press a specific button manually every time your skeleton changes even slightly.)

APEX has the corresponding node (rig::parentblend). It works the same way: the transforms have to be stashed. But in APEX workflow, the whole network is procedurally generated, unlike VOP. So transforms are stashed all the time and it's less an issue.