12. Setup the muscle and phoneme poses: Now you are ready to make all target poses for the expressions. We had control nulls that had attributes for each target- for example, left smirk, left squint, right grimace, etc. Once you have all your controls named you can use Set Driven Key to drive the control faceCurve contol curves in the extremes.


We used values of -10 to 10 (where 0 us neutral) on controls that have two extremes such as LeftBrowMiddle, and 0-10 on standard expressions such as leftSmirk. The process is exactly the same for create phonemes and everything is able to be layered, just like blendshapes. Just make sure you are setting your keys on the proper controls since the upper level control hull will remain available for hand keyframing so the animator can always layer or override the preset targets if necessary.


13. Clean up the face rig: As you can see, the rig is pretty cluttered and you won�t need to see all the individual faceCurve control hulls very often if the target shapes are strong, so you can template and/or hide all of the child control hulls (they are driven by your target attributes so you don�t want an animator accidentally keying them) and then create an attribute to toggle the other controls. We just have a Boolean attribute called �faceRig� and we use Set Driven Key to alternate the visibility attributes on the control hulls on and off.


14. Misc Info: If there are target shapes you just aren�t able to achieve with the curve controls you can always create blendshapes to supplement your rig. Just drive them with another attribute on the controls you use for the other shapes and no one will even realize you are mixing and matching the techniques. We didn�t set any limits on the range of motion on the control curves- I tried a few techiques for this, but they always seemed to somehow limit the animator too much. You can always zero out the control if you don�t like it, so I just gave animators free reign and it was never an issue in our production.

Some additional example movies:
baby delgo.avi
merchant.avi

In closing, as long as you order your deformers correctly, you can add sculpts or any other extra techniques to this rig, but keep in mind, with each thing you do outside of the standard rig described above (which utilizes just one smoothBind cluster node), the less transferable it becomes. This may not be an issue for you, but one of the big benefits we found in the above method was the freedom we had to modify things on the fly. Since we stuck to this method, we were able to make major modifications to meshes (changing topology, UVs, etc) during production and we could always use a simple copyBind to transfer all of this information in a click. The techniques in this article can be used on a lot more than faces- have fun finding ways to implement it!