Facial Rigging: How do you decide what goes to Bones vs. Blendshapes?

Support Animation and Rigging
1

Hey everyone,

I’m working on a facial rig for my short film and I’m hitting a wall regarding the workflow. I’ve been told to use a “hybrid approach,” but I’m struggling to find a clear rule of thumb for when to use bones and when to use blendshapes.
Currently, I have a bone-based setup for eyelid tracking, but I’m wondering if I should be using blendshapes for the actual blinking and other expressions.

Could you help me understand your logic? Specifically:

  1. Bones vs. Blendshapes: Which parts of the face do you strictly keep as bones (e.g., jaw, eye rotation) and which are better off as shape keys (e.g., mouth shapes, wrinkles, blinking)?
  2. The “Hybrid” Logic: Is there a standard workflow? Do you build a base bone rig for mechanics and then add shape keys on top for the “finesse” and deformation corrections?
  3. Efficiency: I want to avoid over-engineering. What’s the most efficient way to achieve a professional look without creating a maintenance nightmare?

I’m aiming for a high-quality result, so any insight into your professional workflow would be hugely appreciated!

2

Here’s what we do at work- jaw bone, eye rotation bones, everything else is shapekeys

3

I’ll preference this with the fact I’ve never actually worked in 3D, but I have spent a lot of time researching, reading, watching and messing around with my own characters.

On that basis, my answer to the question would be that there is no clear rule of thumb.

It would all be very much dependent on the project, the style, the animation requirements, the animation software/system used and likely even the skill of the rigger, etc.

I mean I’ve seen a lot of ‘making of’ and ‘behind the scenes’ of Disney/Pixar films and while it’s possible I’m wrong, I get the feeling that much of it is bones. You can see the sometimes crazy number of bones and controls in the characters they make and I think this is so that have fine control over pretty much every area of the character.

So while its a very high number of bones, it would likely be an insane amount of shape keys (maybe constantly being added during the production) to do the same thing.

On the other hand, say you have a more limited type animation project but a huge amount of frames to animate in a tight deadline (maybe with limited experienced animators).
As such, you spend a bit more time setting up all the required shape keys, at which point you can speed through the animation and it half won’t matter who does it, as the ‘quality’ will look basically the same, since everyone is using the same shape keys.

Personally I’ve so far just used bones and got pretty good results, but I can modify things more and add shape keys later if it turns out I need them.

4

Hello;

It’s a fundamental question. As of 2026, it’s not easy to find in the internet readily accessible guides for understanding the artistic purpose and the exact methodology behind ccertain “standards”. But they sure exist --at least, when it comes to specific contexts of Production. For example,

In a context of a leading Shape Keys approach to Facial Expressions, the reason behind the use of strategic Deform Bones, notably, the case of the big Jaw Deform Bone for Skinning, is not overly complicated, although it’s also not an Initiate thing in Rigging & Animation (although it should!); might even be an Intermediate to Advanced thing to know. Is that, the Range of Motion (especially Open/Close, maybe a bit for Sideways too) of the Jaw, needs to Pivot (somehow); and through the Shape Keys Skinning, it won’t be able to grant, by default at least, spontaneous Pivoting/Rotation for that whole Mesh Region corresponding to all the tissues more directly carried by the Mandible; instead, a Shape Key from A to B, will just offer a “Linear” (literally, a straight line in 3D space) Interpolation for the Morphing/translating Vertices. It seems that even in VRM, for 3D Avatars, involving Facial Mocap (highly dependent on Shape Keys/Blendshapes Facial Rigging), a Jaw Deform Bone can be employed specifically for Skinning. However, there seem to be a more advanced technique, in which, alongside the Jaw Deform Bone Skinning, there is, also, the Jaw Shape Keys Skinning. This double-composed Skinning Setup is not essential, any of the two solutions should work alone more or less fine… the facts are: it would correspond to more work to do; also, Deform Bone Skinning suffers from horrid Skinning Artifacts if the Rotation of the Deforming Mesh (associated with the Rotation of related Deforming Bones) is ‘too high’ (even before it reaches 90° Rotation by the way), though I believe in plenty of 3D Character Designs, the Jaw Mesh tends to survive that appraoch; as for the Shape Key Skinning, it will provide the ideal “Extreme” final shapes for the Jaw Mesh, but by default it just cannot Interpolate Morphing Vertices, from A-to-B, with a curved/non-linear, Rotating/Pivoting Motion, as the Deform Bone does. That is why, the tactic of combining Jaw Deform Bone + Jaw Shape Keys, if done carefully, would tend to provide even better results if needed; of course, Rigging the Jaw with just a Deform Bone I personally believe it’s more straightforward (in the case of use of Jaw Shape Keys, though, Deltas, for Diagonals, as Corrective Shape Keys, might be needed hypothetically; and also tweaking the Drivers curves on their graph…; while in the case of Jaw Deform Bone, sometimes Helper Deform Bones need to be created in order to provide more accuracy to the Armature Deform Skinning approach); however, if you’re looking for very good 3D Character Design for 3D Animation, with ideal Facial Expressions, then maybe you should consider such combination (especially, if you’re already familiar with Facial Shape Keys making-method and are making a lot of Facial Expressions Shape Keys anyways, I guess just 1 more set done shouldn’t be too much health-threatening​:joy_cat:); in theory, it makes sense.

I’ve been doing Facial Rigging Stylized Characters with Armature Deform approach for a while, looking for very good 3D Character Design Facial Expressions… but through an experience of making Rigging & Animation for a 3D Cartoon project recently, I eventually found out what other Riggers & Animators have been trying to tell us over and over again (or maybe, that is just my personal conclusion so far): that the best thing about a Face Rig with Shape Keys approach, is that it makes it so much easier to reach a lot of 3D Character Design quality for Extreme Poses (and also “Key” [in the sense of standard, storytelling] Poses), including certain Exaggerated Facial Expressions. It should remain arguable of course that it would be very possible to achieve such optimal results with mainly Armature Deform Skinning approach… Personally, I won’t personally advise that; besides the background that I’ve mention, these 2 fundamental reasons, regarding Facial Rigging:

:play_button: The best thing about Deformation (Armature Deform Skinning), is that it is good for Tweaking Facial Expressions; However, in a lot of Animated Shows, this “Tweaking” is not as fundamental as we may think; at least, myself, I have been thinking for a long time that it was the most relevant thing; it wasn’t, after I saw what a ‘relatively simple’ 3D Cartoony Shape Keys Face Rig could achieve on a regular basis, on every Animation Shot. That means, if you’re going for Economic/Limited Animation (usually the case fore Series, sometimes Short Feature Films), Tweaking Facial Expressions [with Tweaker Deform Bones] might be totally dumpable; also, because it takes forever to be Tweaking Bones, so it’s just not practical to be Tweaking Bone all the time; of course, in the case of a Long Feature Film, or a high-quality Short Feature Film and etc., then Tweaking can become fundamental I guess. Regarding this Tweaking stuff, it is very important to understand, that the Armature Deform Face Rig, allows for funny (sometimes handy) manipulations of the Mesh which can make “original shapes” on-demand, also favoring perhaps original and perhaps important, specialized Interpolation transitions (this might or may not relate so directly with the Rotating/Pivoting example of the Jaw Deform Bone)… but again, sometimes, these Interpolations are so subtle, that the assumed ‘freedom’ Tweaking can provide, might not really compensate the efforts (of both Rigging and Animating with them); additionally, it is true that, Tweaking can also very easily disrupt the Mesh in horrid ways, and also can cause horrid and unpredictable Mesh distortion during Interpolation events. In that sense, the Shape Keys Face Rig, if carefully achieved and tested (to anticipate unwanted conflitcting Interpolations notably between different Facial Expressions Shape Keys), has a major tendency for ‘stability’, and this, I believe is particularly a great feature to have in a Face Rig, at least when it comes to Economic/Limited 3D Animation projects.
:play_button: What about Pose Assets? When it comes to 3D Animation, on paper, they can probably conjure, a whole Face Rig made out of purely Armature Deform Skinning approach, into something more efficient (like the Shape Keys Skinning approach by default)… So far, that might just work well, as long as Tweaking can be avoided to some extend, in the case of more economical projects.

Personally, I’m sticking with a ‘leading’ Shape Keys Face Rig approach on my projects --meaning I’m looking forward 3D Character Design quality for rich Facial Expressions, while avoiding the hassle of being overly dependent on Armature Deform Skinning for the Face Rig. Probably would sustain a Jaw Deform Bone, and maybe, just maybe, Tweaker Deform Bones (not-leading) for occasions --but that would depend on the result of some experimentation, as Blender is doesn’t provide much support for smarter interactions between Shape Keys and Deform Bones when it comes to the Control Rig (for instance, there is a fundamental Control Rig flaw, actually, when it comes to Tweaking Deform Bones on a Mesh Region which also has Shape Keys on effect).

I hope this can provide some insight and guidelines.

5

All animated face elements are controlled with bones; I only use shapekeys for specific corrective purposes, and very few of them.

6

Hello !

For jaw and eyeballs, bones makes sense.
From there you can really get very similar results with a bone based rig or a shape key based rig.

The good points of bone based rig is that you’re never limited and you can push the rig however you want.
On the other hand, they’re harder to setup, they get very tricky to do very precise shapes and they tend to break easily :
While you can do every shape, you need to tweak the bones a lot to get a good result and make a clean shape. Or you add another layer of corrective blend shape , and it takes even more time to setup.

Personally, I work mostly with blend shapes and it’s much faster to setup. But to make that really efficient I had to make a few python script to automate shape splitting.
For instance you can make a smile shape with brown up. And use a script to split that in two shapes : brown and mouth. Then split each of those in another Left-Right shapes set. Then automatically add drivers to link those shapes with bones.
Then in fact you can split even more, like having each brown split in 3 parts to make even more variations…

When you have scripts like that , you can rig very fast and have very well defined expressions which are very hard to get with bones.

But… there is no strict rules, you may want to use shapes only for the mouth and do everything else with bones… Why not and it depends on the case…

I think if you’re getting started you can either go with the simplest approach for each element (eyelids, browns, mouth) or test different approach before making your mind…
The rule of thumb is that, given you have scripts, shapes will be faster and when you have a lot of character with different proportions it’s easier to re-purpose the same rig and shapes will always work no matter how different characters are from each others.

But, if you have time and not a lot of characters you can use bones and it’s fine.
It’s also possible to have a somewhat imperfect ( = fast to make) bone based rig, and use pose library to save a set of facial expressions. So even if the rig is a bit bad to operate, you can work fast with it using the poses as starting point.

As rigger it’s not set in stone which approach you’re going to use. It varies given what the rig is expected to do, and how much time you have to pull it out.

If you have a screen shot I can tell you what I would have done in that particular case.

I also recommend stop staring book from Jason Osipa that explain very well how to rig a face with shapes and I think they also cover a bit of bone based rig in latest versions…

Have fun !

7

Welcome! :vulcan_salute: I’m not a professional, but I’ve managed to do pretty good facial animations with only bones. When I first started doing facial animations, I used only shape keys, because they were faster to create than trying to learn facial rigging and facial weight painting. But after a while I switched to using bones, because with shape keys I had to start creating them from scratch every time I made a new character. Also shape keys make everything messy because you cannot apply modifiers or make topological changes any more (at least with vanilla Blender). But with bones I could create pose assets and use them for every character after that. And as I’m using the same rig for all characters and always start with a fully textured and rigged base character, I can work very fast. Sometimes I can even create several new characters per day. But, ofc my approach has its limitations, especially with talking animations. Vowels a, e, i and o are easy to make with just bones. But more extreme expressions like u or a kissing lips pose are impossible to get right with my rig. I’m yet to make my first animated film, but as I do it, I’ll try to make as much of it with bones. But when they fail, I’ll use corrective shape keys. So the approach you choose will depend on your personal needs and preferences. :+1:

8

That makes sense ! Indeed if your characters have always similar proportion it might be worth building one good bone based rig.
That also depends on how far you want to take things, how much automation and so on…

Getting refined facial expression with bones is doable but much harder :

Also, I don’t know what people think when they do shape keys. But say having shapes for M,B,P,O and / or expressions like happy/sad is kinda the old school method. Nowadays the face is broke down in much simpler shapes and the goal is to mix them to form a more complex expression.

This is taken from Osipa’s book which might be the first to develop that method and explaining it in their book :

image
image466×790 65.3 KB

The face comes with a limited set of controllers but they mix in a logical way.

Again, you can do that with bones, but it’s much more time consuming.
Probably it’s also possible to do mixed rig, so you borrow a bit the best and the worse of both world. Broad movements with bones, small details with corrective shapes.
The overall benefit would be that the rig is going to be more flexible and could do more extreme pose in case of cartoony rigs !

9

Yeah, because I’m doing realistic(ish) characters they tend to all have similar proportions :smile: stylized characters have more variation and more exaggerated and different shapes.

I have “locked” the position of the mouth and eyes for all my characters. Eyes because the eye bones need to be in the center of the eyeball for them to roll properly. And mouth because it’s the part that most likely will have interactions with objects and characters (drinking, eating, kissing etc). If I made more animations including aiming with one eye then it would also be the eyes. For game characters I never change the rig so that I can use the same rig for all characters. But inside Blender, if the character really needs to have eyes in a different position, I’ll just make a custom rig and adjust the eye bones accordingly. The eye bones track to a sight target bone, so moving them will not mess up the directions the characters are looking at. In my experience eyebrows, cheeks etc deform pretty well even if the bones are not exactly at their positions. So I’ll only adjust the eye bones and eyelids surrounding them, as they are more sensitive to weird deformations. :+1: