With each passing year, more and more CG-driven films and television programs are showing up on screens around the world. Every so often, breakthrough projects scale new technical heights with increasingly impressive performances by their computerized characters. No matter how far technology advances, recent successes such as those achieved by WETA and their magnificent work on “The Lord of the Rings” trilogy prove that these characters can transcend their seemingly limited worlds of wire frames and renderings to become living, breathing on-screen creations. From concept sketches, to sculpted characters, to fully rigged models, an enormous amount of teamwork and effort is involved in developing the digital foundations of these characters. One of the main keys to the success of these efforts is the ability to create convincing emotion on the faces of the characters they develop.

Nearly every ‘behind the scenes’ segment for a CG film contains some quick snippets of an animator manipulating the facial expressions of a CG character. Often described as “special effects magic” these segments usually reveal about as much of how the scenes were created as traditional magicians will reveal of their secrets. In other words–nothing. Let’s pull back the curtain a bit and learn something about how these digital artists breathe life into their inspired cinematic creations.

One of the most essential elements in the creation of CG characters is the development of facial poses using blend shape targets. Most production houses have created their own proprietary animation controls and deformer systems for producing and animating these different target poses. Fortunately, you don’t have to have a job with Peter Jackson or George Lucas to get access to this kind of top notch technology. Alias’s Maya provides both intermediate and expert users the means of achieving incredibly professional results with its own blend shape editing system. This system is fairly easy to understand and use, let’s discuss some of the basics.

Maya’s blend shapes are employed to ‘morph’ one facial expression into another during animation, allowing you to combine different base expressions to create the complex emotions and facial positioning that are intrinsic to emotion and speech. Essentially the name says it all since what they do is seamlessly blend the geometry of one surface or model with another, much like a Deejay will combine musical elements into a new compilation.

1: The blend shape editor along with the neutral base head on the right and three target heads on the left.

Maya also includes an editing interface to not only control shapes and surfaces, but to modify and create new ones, as well. Appropriately enough, this tool is called the, Blend Shape Editor Blend shapes can be created from almost any type of geometry including polygons meshes, single surface NURBS patch surfaces, and multi-patch NURBS surfaces. There are definite advantages to using subdivision surfaces geometry when employing blend shapes. Geometry created in this way develops via a hierarchal system of detail in which the higher levels are driven by the lower. In plain English, this means when you animate a sub-d model at its lowest level of detail, everything at a higher level of detail will follow. The head models in the illustrations accompanying this article were created using Maya’s subdivision surfaces.

The first step involved in setting up blend shapes is to select a base head or face with a neutral or blank expression. This head can be one that you have already created, or you can model a new one from scratch. The next step is to create a series of target heads, each representing a different facial pose or expression that you want your character to communicate. This can be done by simply duplicating your head model and re-sculpting its features into a virtual library of different expressions. The more you create the more creative options you’ll have when animating the character later, but remember–different blend shapes can be combined and edited to create entirely new targets without any additional re-modeling, so don’t go overboard.

Important: After creating a new blend shape, be careful not to add or delete any of the head’s editable component geometry such as edges, faces, or vertices. Doing so can create radically contradictory deformations which usually turn out to be an irreparable mess!

Also watch out for how the geometry is structured since geometry that is too complex or placed in the wrong places can create unwanted deformations during animation.

The best way to approach the creation of your blend shapes is to make sure you’re really happy with the structure of your base model before creating the targets.

The core of successful facial animation using blend shapes is convincing and expressive target faces. All of the target faces one can model are based directly on the underlying musculature of the face and their dynamic interaction. These faces can be created by simply editing the surface geometry of the base model to create various expressions, of surprise, joy, anger, etc. More sophisticated facial animation calls for the employment of facial targets know as phonemes. These poses represent the way in which the features of the face move and deform when making a particular sound or basic components of speech such as those made when speaking vowel and consonant sounds. Let’s explore some of the more common facial targets.

Tip: Use a mirror for reference. There’s no substitute for looking to nature when trying to create naturalistic works of art, CG or otherwise.

Tip: If you’re working on a slower or older system, one trick is to parent your hi-res model to a lighter, low-res version of itself and perform all of your animating on this model. This will allow you to work a lot faster.

2: Figure 1-3. Phoneme target for the sound, “T”.

NOTE: For this part of the work, make sure Maya is set to its Modeling view mode, and that each new model is created from a duplicate of the base model.

MODELING SAMPLE PHONEMES One of the most commonly employed hard consonants used in phoneme target modeling is that of the letter “T”. Emulating the natural movements engaged in the creation of this sound requires more than simply manipulating the vertices that control the surface of the skin. Because the mouth opens slightly when producing this sound, both the teeth and lower jaw will need to be adjusted. First, make sure that the lower jaw and the teeth are grouped together. Looking through the side view pane, drop and rotate the group to open the mouth slightly. Do this with X-Ray shading enabled to reveal a complete view of the interior structures of the mouth (Shading>Shade Options> X-Ray). You’ll find it’s easier to rotate the jaw if you move the pivot point of the lower jaw group to the back of the mouth. This helps to mimic the movement of the hinge joint of the lower jaw when using the rotate manipulator. Do this by selecting the group and hitting the Insert key to enable and adjust the position of the pivot point. When you’re satisfied with the position of the pivot point, hit the Insert key again to lock its position.

Proceed to drop the lower jaw group down a bit by using its move manipulator. Next, open the jaw by using the rotate manipulator to tilt the front end downward until you’re happy with its position.

Now let’s move on to the lips. If you haven’t already done so, create a layer for the lower jaw and all of the teeth it contains. Set the display type of this layer to “R”, for reference. Choosing this display type establishes the objects contained in the layer as reference objects which cannot be edited. Now begin to modify the opening of the mouth to mimic the structure of the lips when speaking. Raise the upper lip just enough to show the tips of the upper front teeth. The lower lip should drop slightly to reveal the tops of the bottom front teeth in their now open position. The trick here is to capture the subtlety of how the fleshy portions of the mouth deform without lapsing into caricature. Once you’re satisfied with the position of the lips, the next step is to move the tip of the tongue so it’s just pressing against the backside of the upper front row of teeth. For those of you who enjoy adding a little nuance to your modeling, pull in the cheeks just a touch to imitate the slight depression formed when the mouth opens.

Tip: Keeping the components of the head in separate layers prevents the accidental modification of the parts of the model you wish to be left unaltered.

3: Positioning the tip of the tongue behind the upper front teeth using X-Ray shading in the side view pane.

Another consonant common to speech is “F”. Creating this target will also require opening the mouth a little. This time however, instead of repositioning the tongue, it’s the lips that will require the most editing. We’ll start this exercise by working on the lower lip of a new copy of our base model. Select and manipulate the vertices of the front part of the lower lip and pull them back so the lip rests behind the upper front row of teeth. It should look as if these teeth are almost biting down on the lip itself.

4: Phoneme target for the sound, “F”.

Take a look in the mirror and pay attention to how your mouth moves when you pronounce the sound, “fff”. You might notice that there’s a slight “over-bite” effect created you do this. This is caused by the way the lower jaw shifts slightly backwards. Try to create the same effect in your model by moving the group of the lower jaw and teeth back using the move manipulator.

The next thing to do is raise the middle of the upper lip so that some of the upper front teeth are exposed. Do this the same way you remodeled the upper lip on the “T” target head with the exception of exposing a little more of the upper front teeth. To help create the illusion of air being pushed through the mouth as this sound is made, curl back the top portion of the upper lip somewhat. You can also pull out the corners of the mouth to make it look like the skin there is being acted on by air inside. Again, try to accentuate the intended effect without making the head look cartoon like.

5: Curl back the top of the upper lip to foster the illusion of air being pushed through the lips.

Tip: If you are using Subdivision Surfaces to create these models, employ simple vertex editing in Polygon Mode for quick results. Model them in Standard Mode when adding effects requiring more subtlety.

6: Editing a Subdivision Surfaces Model in Polygon mode.

CREATING BLEND SHAPES Now that you’ve created a base model and a few target heads, it’s time to move on to the blend shape set up. After you’re fully satisfied with the base and targets that you have created, change Maya to display in its Animation mode. Marquee-select the modified heads and shift-select the original, neutral base head. From the menu bar, select Deform> Create Blend Shape. This links up your target heads which can now be controlled with the Blend Shape Editor. There are an enormous variety of facial expressions and poses one can create using this feature. Each target head’s expression is represented by a deformer control slider in the Blend Shape editor’s window. The parameters of their respective deformations can be adjusted either numerically or via the sliders from a default weight of 0.000 (neutral) to a maximum of 1.000 (the full extent to which the target head has been modified to deviate from the original).

TIP: The range of the Blendshape slider can be extended by typing a number outside of the 0-1 range in the edit box. Numbers beyond 1.0 can be very helpful for some poses, as can negative values below zero. This ability to create additional ranges of “free” model poses is very helpful, but care must be taken to make sure unexpected problems do not arise from these non-modeled targets. This can happen when animating and often results in teeth flying through upper lips!

Experiment with manipulating the various sliders to transform the base head into each of the various target heads you’ve modeled. Next, begin to move the sliders together to create entirely new poses resulting from the combination of the various target weights. As you can probably imagine, an almost unlimited range of expressions can be created by combining this feature with a handful of properly constructed models.

7: Moving a combination of target weights to create a smile on the model.

With experience and practice you’ll quickly develop a strong command of theses operations. This new understanding will set you well on your way to developing convincingly expressive characters for your various CG projects and animations. As mentioned above, modeling target faces is just one small part of the enormous creative effort employed to bring today’s most memorable CG creations to life. At least you’ll never have to scratch your head in wonderment the next time you watch the behind the scenes section of your favorite DVD. By that time you’ll have already learned that the secret behind today’s “movie making magic” is nothing more than old fashioned hard work and dedication. Who knows, maybe that dream job isn’t such a crazy idea after all?