Website Youtube GitHub

mGear Framework Forum

Gimmick Joints vs Blendshapes

Why should i use Gimmick Joints and dont go straight to Blendshapes?
Im a Modeler so i would just use Blendshapes.
In wich case are Gimmick joints the better solution?

I assume a combo of both would give me the beste result?

Hi again @oglu

Some people like some techniques better than others. Both comes with good and bad things.

I always go to blendshapes because I think you can achieve more detailed deformation (and anatomy preservation of the asset characteristics) when sculpting shapes than with joints but I have been proven otherwise in the past… I just prefer blend shapes because if you want to achieve the same level of details on complex assets with just joint base things you end up compexifying the rigs.

That been said you need to know that the issue with blendshapes is as well how things get driven and how sculpts combine themselves. RBF driven nodes are not always understood by riggers and same goes for cone base ones because sometimes riggers themselves fail miserably explaining them.

The only constructive advise I have on this is for you to handle this with a technique you feel comfortable with but be sure you understand all the implications of your choice.

Myself I only use joints base corrections in situations where I can’t drive the correction with a single transform/joint.


Hi Jerome that is great info.
I will go with Blendshapes first.

gimmick joints (intermediate joints) is a very old technique (before dual quartenion exists) to preserve volume and getting good deformation around a pivot…
the advantage is, that you dont have to provide for every rotation axis a blendshape with driven keys, wich when it comes to gimbal can be very painfull to fix…
(i.e. in shoulder area you could make for every loop a gimmick-joint with distributed weights to get a nice smooth deformation…also helpfull in ellbow, hip, knee areas or spine…)

1 Like

@Oglu, just adding to what @Jerome said, for some cases using gimmick joints will be the right thing to do. For example, in our case, where we have one mesh and hundreds of possible garments, it’s faster to project and clean the skinCluster than projecting and cleaning corrective blendshapes in all the garments.

As an example, at Disney we used a combination of skinCluster + wires + cluster + driven keys as a basic layer of deformation, and we added blendshapes on top (sometimes, ridiculous amounts) to have a better control of how things look. Also, remember that when you model shapes for a character you not only think in terms of appeal or volumes, but also about motion: you can have a beautiful shape as a final result, but if the interpolation between the rest pose and that shape, or the rest of the shapes and that one, is terrible, you are going to have a bad time dealing with the whole system.

Like anything, the more practice, the better results!


Thanks for the detailed info. The part with the practice is the main problem. I have zero.