im curious about the general consensus is regarding different facial set up’s.
i know joint based facial rigs are common in the game industry but is it the same in films?
Most of the popular game engines support both blends and joints but from what I’ve seen usually developers will choose to use one or the other. There are a few we’ve worked with that allow both and in our experience that is what gives us the best result, a hybrid joint/blend facial rig. This is usually in the form of a joint-driven setup with corrective blend-shapes applied where needed. Sometimes we’ll add blend-shapes for very specific shapes that are difficult to hit with joints if the budget allows it, such as sticky lips.
In film, in almost every case you’ll be working with a blend-shape rig, but we’ve done some work recently where we’ve been building joint-based hybrid rigs where we’ve achieved quality on par with full blend setups. The benefit to this method is that it’s much easier to work with and gives the artists infinitely more possibilities for shapes. Our animation team for the most part can’t stand blendshape rigs due to the limitations presented by the specificity of the shapes. Faces do really odd things sometimes, and you need the ability to hit those shapes when you need to.
To answer your question simply, no…joint based facial rigs are not common in film, but they do exist. 
I come from an animation background initially and I can vouch for having bones in a film rig, particularly for lip rolls, blends are very linear when they transition, the mouth and jaw have much more going on in the way of arcs.
So as Jay says animators do prefer them, freedom is key.
[QUOTE=JayG;7439]
…In film, in almost every case you’ll be working with a blend-shape rig…
[/QUOTE]
Been using joint based facial rigs in film work since 2006. Haven’t touched a blend shape for anything more than a corrective since.
These days however it’s a combination of bunch of custom deformers really…
[QUOTE=ShadowM8;7442]Been using joint based facial rigs in film work since 2006.[/QUOTE]
In that case, I retract that part of my statement, and…awesome! Glad to see others doing it.
We here use a joint based setup as well, with blendshapes being used for super extreme poses we need that can’t be achieved with a regular setup.
[QUOTE=ShadowM8;7442] These days however it’s a combination of bunch of custom deformers really…[/QUOTE]
thats very interesting actually, can you elaborate a little bit please?
do you mean you guys duplicate curves off the face and set weights to wire deformers?
Yes, for example on Tangled at Disney we relied heavily on the use of wire like deformers (they were our own improved versions but principle is the same). We had other surface sliding and collision deforemrs we use all over too. Custom shink wrap like deformers were used as well in many cases.
That being said a rig like Benjamin Button is all blend shapes at it’s core.
[QUOTE=ShadowM8;7450]Yes, for example on Tangled at Disney we relied heavily on the use of wire like deformers (they were our own improved versions but principle is the same). We had other surface sliding and collision deforemrs we use all over too. Custom shink wrap like deformers were used as well in many cases.
That being said a rig like Benjamin Button is all blend shapes at it’s core.[/QUOTE]
whoa! very cool!!
thanks for the insight :):
im putting together a quick reel to show some of my latest stuff to a recruiter from DD Port st Lucie for a possible internship up there. i think this these tips are gonna help me out a bunch!!
thank you!
I had no idea joint based rigs were so common. I’d like to investigate ways to make joint based facial rigs work out better because here at Double Fine the characters are so stylized that I had trouble producing decent results with just joints. On Brutal Legend we ended up using blend shapes for everything. Not phonemes though. Each shape would simulate different facial muscle combinations. We got great results and variation with that system, and it was simple enough to keep animators building characters and animating at the speed we needed, but obviously if you can get joints looking good it would provide much finer control.
The hybrid system mentioned before sounds like a solid direction to go in.
For face rigging to work with joints, like most good rigging, a solid grounding and understanding of correct joint positions and then a firm grasp of weight painting is essential. Layer over the top of that a deformer like stretch mesh or some relax deformers to help with the areas between the joints and you can do some excellent rigs.
The decision to mix them or go one vs. the other comes down to understanding both systems limitations and your overall pipeline…do you have a large set of modelers and an established blendshape pipeline/ VFX shapes being created from actor face scans etc… or are you going to be in a free form animation environment where animation control over every attribute is needed and you may have less time to build shapes or less modelers.
There are many hybrid approaches that work really well as mentioned here.
I just wanted to make sure that others reading this thread realize there are many factors that will shape the decision for joints or blendshapes and both are valid approaches.
From the perspective of game development…
Shapes
Pros:
Can give effect of muscle movement under skin
Relatively easy to use in a motion capture pipeline, provided your solver is set up right.
Shapes, and therefore expressions are explicitly controlled by the character artist. Dialling in a certain expression will give you exactly what they intended.
cons:
Linear vertex movement
You need quite a few shapes.
Can be difficult to get a 50 shape system working together and requires a lot of modelling resources.
Joints
pros:
Non-linear vertex movement
cons:
Can be tough to get nice skin weighting where you are limited in the number of joint influences per vert the engine will support.
Higher animation data cost.
Can’t simulate the effect of muscles sliding under the skin.
The nature of using joints and skinning to control facial expressions means it’s impossible to precisely hit the expression as envisaged by the character artist, there’s an extra layer of interpretation that doesn’t allow you to achieve absolute shape prescision. It’s probably not an issue for stylised facial animation, but it’s a massive problem if you’re trying to capture the subtleties of human expression.
What I do sounds something like the hybrid system mentioned, except I call them active blend shapes. They’re blend shapes out in space that have joints or some deformer on them (I prefer clusters - they’re less messy than joints). The deformers are then keyed to the controls on the face or off to the side on a menu system, causing the blend shape mesh to deform. That deformation is then sent to the tunneler, then to the base mesh on the rig. I find that in general this leads to a smaller, faster file and allows for a true arc within the deformation.
We use a lot of blendshapes.
For claymation type characters, it’s pretty much a no brainer. Basic joints for legs and stuff, and plenty of shapes for all the poses. Each extreme then morphs the joint rig into an appropriate position, since they’d otherwise be dis-aligned with the extreme new pose of the mesh.
Facial rigs get blendshapes aswell, altough I can definately see the merit of using joints aswell.