Home page > Animating Willy the web wizard

This article covers a quick run through of the process taken to produce a collection of short dialogue animation clips for Icanect and their online web wizard for creating custom web page templates. Thanks must go to Bob Hurwitz for allowing these sequences to be used publicly. Take a look at the below avi files (approx 5hrs work per movie, files are Mac and Windows compatible). I don't pretend to be a god like being of the animation world but if you want to know a little more as to how they were done, read on...

Phoneme shapes

The mouth rig uses a pose layering technique for greater control using pose sliders, driving individual muscle groups rather than a complete phoneme target. These muscle group rigs initially make it more difficult to define the phoneme shapes, but once done, they provide more flexibility during the animation process, especially when layering a more expressive character performance together.

Eight pose sliders are enough for a basic lip synch: Lip Stretcher, Lip Tightener, Sneering Muscle, Bottom Lip, Tongue, Smiling Muscle, Sad Muscle and Jaw.

Each of the basic phoneme shapes shown below can be formed by setting just these eight pose sliders. A simple table made from of each of the eight values required for each of the major phoneme shapes can be used as reference during later lip synch key-framing.

Blocking out lip synch keys

The audio tracks were provided by the team at Icanect, each track was imported and dropped into an Action ready for use. Before starting pose sliding, I recommend opening up the model Pose container (in the Project Workspace) and selecting each of the poses in turn, checking or un-checking "Display slider on Pose Sliders window" property - you only need to have the mouth sliders displayed so you can work uncluttered. With an open Action window, holding the dialogue, and the Pose Slider window ready, its time for the tedious bit...

My method of phoneme blocking is to start by setting keys on all of the eight pose slider values at each needed phoneme, stepping through the Action frame by frame, dropping keys where necessary (using the Pose Sliders numerical inputs from the reference phoneme shape value table). For mouth shape holds, I prefer to just drop in extra keys, rather than breaking the work flow by adjusting channel curve magnitudes - do that later ;o). As to which phoneme shapes are needed at each sound, practice makes perfect, it's not a difficult as it may seem at first...

Once the first pass of the dialogue track is complete, it can be rendered out and played in synch to the dialogue sound track for closer analysis. It's a simple process to then open up the eight mouth slider channels and adjust timing and add extra keys where needed. The wizard sequences as shown above took between 40min and 3 hours each to key the basic phoneme shapes and tweak the timing. Shown below is a channel view of the lip synch data for the wizard_wrong.avi.

Body movement

Ok, Ok... I admit it, I did this the wrong way round!! It's generally agreed that you should animate body movement first and lip synch second. This helps you to focus on making the body communicate what's being said before you start playing with phoneme shapes. Next time, I promise. ;o)

I wont go into too much detail as it would get wordy, but here are some basic tricks I used during body animation:

• Try and animate from the bottom of the bone hierarchy up, ie. Go through and animated the pelvis first, then the torso, then the head, and then the hand nulls (IK targets for the arms). If you don't you'll find you'll have to go back again and again re-keying.
• To reduce the number of items that need to be animated, a number of "Aim At" and "Orient Like" constraints were used on the beard and pointy hat bones. These try to keep their centre of gravity in roughly the same location while allowing some movement as the character body moves about. These days you might want to check out the alternatives: dynamic constraints, lagged constraints, or soft-bodies.
• For a little more snap to the movement, adjusted the channel curves for the head and hands so the magnitude of the curve into a key was smaller than the magnitude out, more of a "hit the pose fast, hold and ease out".
• Blinks really help keep things alive, it's just one pose slider channel key with -100% bulging and +100% squashed. The keys ran something like: 0%, -100%, 100%, a few frames, 100%, -100%, 0%. This gives them that anticipation bulge just before they close and the overshoot as they open again.
• With a character talking to screen, added a slight eye twitch; switch his look between where your left and right eye would be. This is just a matter of keying the eye null target to jump between two positions near the camera.
• Once happy with the sequence, well maybe happy isn't the right word... Ok, once you need to move forward with the sequence, go through your keys in the channel curve editor and remove twinning (where your character hits a pose all at once). Also add some timing overlap to help the unfolding of poses; usually (depends on where the motivating force is), start at the bottom of the bone hierarchy, move these lower items earlier in time and items higher later. Warning, it's best to do this once all your timing is solid as your keys will look a real mess after this stage - you don't want to come back to make major edits after this.

Well that's about the worst of it, glad you read down this far, I at least hope it was of some help. If you haven't seen them yet and want to see more of web wizard Willy there are larger renders on the other web wizard page. Below are a couple of extra screen shots showing the character's simple mesh and bones.

Related links

• Extended phoneme reference mouth shapes
• Preston Blair reference mouth shapes
• VST plugin wizard vocal musical interface element
• Web wizard character design
• Interface wizard a character interface element

Back to Gary C. Martin's home page