Well... where to begin... there is only one (1!) example of this on the internet, and it's for Maya in the old Jason Schleifer's "Animator Friendly Rigging" tutorial... and it's a s**tstorm of epic proportions... separate window with multiple choices options, on the level of a plugin written in MEL... yes it works, but I don't want to pick bones and controllers in the hierarchy I want it automated... one button with HDA's python script... Then there is, again, one (1!) Sidefx rigging tutorial series where this is shown but (you can call me stupid) I didn't understand how it is supposed to work... btw there were some complaints that this setup wasn't working properly because of some bugs in at the time version of Houdini... When you look at the hierarchy of reverse foot you have 10 (or eight depends on complexity) up to 12 objects who control two bones rotations... opposed to only two FK controllers (ankle and ball/toe wiggle). Thanks to Toadstorm (cheers man), I know how to handle/dismantle/patch/ matrices to get rotations/translations and matching FK to IK is not a problem... but how the hell am I supposed to match combinations of rotations of 10 or 12 objects to only two objects (matrices)that I can access? Should I create/add additional (fake) objects/nulls in the FK hierarchy who will somehow (don't know how) provide data needed for matching? Help...

Something like this?

https://www.sidefx.com/docs/houdini/nodes/sop/kinefx--reversefoot.html [www.sidefx.com]

reverse foot matching may be tricky and highly depends on your reverse foot implementation
also I don't see a large benefit of matching reverse foot controllers to fk bone rotations
first of all you would need to be restricting fk to even be able to match it
and second of all the matched pose even if perfect may not be what you expect as the tip angle may be arbitrary or always 0 since its also possible to align all foot instead for example

I feel like partial matching and then bending is much more useful approach, but I guess it's up to an interpretation for the rigger on how to approach it and what's the best solution for particular setup

@Jonathan

I'm trying to build rigs in object level... bones etc. Somehow Kinefx feels more as a solution for mocap as far as I can tell... I'm getting some flashbacks of Motionbuilder approach when I see it... Thank you for suggestion and time...

@Tamte

Yes it's tricky as hell... I don't know even where to start, or which object to align first... and to what! Benefit is much easier animation if a character jumps... when I remember situations of dragging IK legs trough air while the character is airborne... tears start flowing by themselves... By "partial matching" do you mean matching up to the ankle (hip->knee->ankle... like an arm), and the foot, with it's infamous footroll, is a separate "entity" or "module" who follows IK or FK chains as needed? Some form of broken hierarchy with a null who acts as a parent for footroll, but switches it's parents (i.e. IK or FK bones) trough parent constraints? Sounds doable... maybe...

@kriegmesser sidefx is actively steering away from object level rigs, reasons for this are quite well explained on the 18.5 kinefx reveal video. Currently it lacks the high level animation tools so it's indeed better suited for mocap processing at the moment. They hinted the animation bits and tools are being worked on, hopefully some of that will be in H19. So while in its still somewhat in its infancy kinefx is the future of the character pipeline in Houdini.

(and been using it quite a lot for hybrid sim'ed/anim/mocap rigs and it's all quite nice once you get the hang of it, but you need to build your own anim tools for now)

So the spline+points = bone skeletons are the herniated future of rigging in Houdini? God/s help us... If they want mocap they should just copy Motionbuilder... Or focus more on interoperability... For now I would love stable/unshakable/monolithic/ parenting function... I've seen constraints that fail in Maya... or wiring parameters that go haywire in 3DS Max... but parenting or parent constraints rarely (never) go down...

It's joint based rigging, so each point is actually a joint, the lines are just there to visualize the hierarchy. So there are no bones at all, similar to Maya rigs in that respect.

Anyways, joints in SOP is where it's at since bones in object level didn't work out that well for a lot of messy reasons.