2d head (that always faces camera) on 3d body


Thinking about sticking 2d alpha footage of talking head on 3d body. I expect it to look ridiculous, but I have only vague awareness of weighting issues (using RH tools), and wondering if having the head always face the camera is as simple as maybe having an invisible object for the head that’s part of the weighted chain, then creating a plane object with a face camera tag that’s a child of the invisible head?

Something else?


Might check out CV-AR

Only works with iPhone X and later.


Looks cool. Hadn’t seen that one. I looked into Crazytalk, and some other similar solutions for bridging the 2d/3d divide.

I’ve got a unique situation in this case, though, where I have a non-standard face shape tha’s already animated in 2d.


I have an older version of Crazy Talk. Creating a 2-d animation that you add as a texture to your 3d face might be good solution. It certainly would be simpler.

Simpler…If you can avoid re-mapping 3d geo.


It’s a neat trick in general, but there are all sorts of specifics in this example that would make any sort of 3d transfer difficult to impossible (non-standard face at weird angles with unrealistic changing geometry, etc)

That’s why all I’m trying to do in this case is to have a literal 2d head on a 3d body.

I’m just wondering best way to make it so face always moves with body, but always faces camera.

Would that be as simple as using standard 3d character, then making head invisible, crating invisible plane as child of head just in front of neck, and projecting the face onto it with face camera tag?

Or maybe some use of frontal projection?


Without some sort of reference it’s difficult for me to understand what you are trying to do exactly.

I’m having a difficult time imagining how a 2d head will look correct or effective on a 3d body.

One could put the 2d texture on a plane (w/alpha of course) and just parent it to the 3d neck, and adding a point to camera tag. I’m just very unsure about how that will look. Could be fun and fanciful…quirky in a good way. Could also look ridiculous.


I expect it to look ridiculous.

Just wasn’t sure if I was missing something about weighting within rig or something like that rendering the point to camera tag inoperable. I just vaguely remember last time i worked with that tag running into problems depending on whether the child or parent of tagged object was being controlled by something else.