Stop Staring – Facial Modeling and Animation Done Right

By Jason Osipa

Sybex, 2003

7½" x 9", 336 pages

Here is a book that I've been wanting to read for a long time but every time I picked it up and browsed it, I found very interesting bits of information both in text form and in image form but overall, it always seemed too technical and too slanted toward rigging and since I was not already involved into rigging, I just put it back on the shelves for later. Reading the book did not change this first impression a lot. This is mostly a book for someone who plan to rig a face for animation or in the case of Animation:Master users, for someone who plan on installing the FACE interface on a model. This is clearly for advanced modelers and riggers although a lot of the content would be very usefull to know to anyone before starting to model a face.

From the point of view of an A:M animator, this book can be divided into 3 types of interests: 1) general morphology of faces, 2) general facial animation techniques and 3) technical instructions for users of another 3D package.

This is little known but Jason Osipa is an A:M user and some of the animations for his two toon characters that are found in the book can be found on A:M film here. There are a couple mentions of Animation Master in the book but that is basically all. The last chapter of the book uses the two characters that were designed and animated in Animation:Master but there are no mentions of how the face splining was designed, how the facial rigging was done and controled and the model files are not on the CD-ROM either. That was a little bit of disapointment.

The first 4 chapters of the book, about 70 pages, is the most general section and is filled with interesting informations and fine observations about the face, how it looks, why it looks the way it looks, etc. I found this section the most interesting one in the whole book. Mr Osipa not only points at often overlooked aspects of facial expressions but also debunks a few common misconceptions.

I would have liked if he whent a little bit further in debunking some misconceptions though. For instance, he debunks the general misconception that brows arching (in and out) are important in facial expresion. While doing that, he mentions the effect of head tilting in the appearance of the brows. I think that this sections would have benefited in more explanation and examples. One error that I often see in head and face models is to build the model with slanted eyes and frown eyebrows not taking into account that this appearance is caused by the head being photograph tilted forward. The resulting model becomes impossible to animate into any other expression than an evil one. I think this should have been covered in much more details. In essence, I found this section too short. I wanted more.

Nevertheless, this is a good section and the stuff found there can be used by anybody whatever the used animation package. The facial landmarking is really nice and shows how important are the folds and wrinkles inn the face to amplify the expressions. It also contains a chapters on animating the mouth that explains Osipa's lipsync method in different steps. This uses a very generic mouth shape to help make a point of what is really important in lipsyncing. His notion of visime goes against the common notion of phonemes championed by Preston Blair, that is used in traditional animation but from my own lipsyncing experience, this approach is better than the phoneme approach.

Chapter 5 on constructing the mouth contains construction techniques that can be applied as is in A:M like the idea of building from concentric spline loops, what Osipa calls the point layout. However, that chapter is so filled with specific constructing instruction using nurbs and polygons and converting from one to anothert and using the appropriate tool and technique which gets so convoluted at some point that it gets boring pretty quickly. Doing that sort of modeling is so much easier in A:M.

Chapter 6 on seting the key shapes is a littel bit more directly applicable to A:M once we skip the long complex explanation on how to do and use morph targets. Here again, A:M poses seems glaringly simpler compared to the intricacies of managing those targets. But then, the explanation of the different facial poses and how they interrelates to form expression is totally applicable to A:M. Every basic facial pose is illustrated with a weight map to illustrate how the affect the different parts of the face. each aspect of each pose is covered including height, width and depth and tehre are even coverage of some of the most common errors in building those poses. Very well done.

Chapter 7 covers the fundamentals of the emotion expression through the eyes. All the chapter is very general as it uses a gery generic and graphically simplified face to drive some points. Everything in there is very valuable information that can be applied in any animation occasions. It contains a few animation example based on sound tracks where Osipa emphasizes some of his animation system (system meaning set of animation rules).

Chapter 8 builds the eyes and brows. Apart from a few general modeling principles that applies to any modeling paradigm, this chapter is mostly filled with specific instructions not related to A:M. Unfortunately, in order to read the pieces of advices that are generally applicable, one need to read all this intricate set of instruction.

BTW, Osipa suggests that strange idea of rotating the eyeballs some 9 degrees toward the outside. That is a weird design decision as it makes characters suffering from outward strabism where the viewer can never quite decide where the caracter is actually looking. This is quite obvious in the face rendered in the book. I don't know for sure where that would comes from but I designed and directed the production of an ophtalmology multimedia CD-ROM several years ago and I never saw any data or evidence that would support that. The only fact that may explain such a decision is the position of the blind spot where the optic nerve connects to the eyeball, which happens to be at about 20 degrees outside which means that if one was to measure the angle between the center of the iris and the center of the optic nerve, that would give something around 9 or 10 degrees. But if this is what motivated Osipa's decision to rotate the eyeball, then that is a big misunderstanding because the reason why the blind spot is 20 degrees outside is exactly because the eye axis is oriented very straight forward.

Chapter 9 is more interesting as it covers the eyes and brows fundamental poses (called keys). There again, one must skip the load of technical instruction that don't apply to A:M (as an example, there are 3 pages of instruction for adding a bumpmap for the forehead wrinkles where the bump percentage can be animated. 3 pages! At some point it looks archaic and becomes funny). BUt the more general instruction on the poses themselves and how they will interrelate together is very usefull and can be applied to A:M poses just as well.

Chapters 10, 11 and 12 are of very little relevancy to A:M It is actually quite a dull read.

Chapter 13 is the last chapter of the book where Osipa shows step by step development of a few animations based on some sound tracks using his facial rigs. It is interesting to follow the author in his animation and read what motivated his decisions at every step. Jason Osipa obviously, have a very analytical mind. It looks like every slider he uses in an animation is motivated by some rule. Rules of the form "When the voice sound does this then I usually make the facial animation do that". At one point in the book, I had the idea of collecting all those rules and edit an "Osipa's facial animation system". But I don't think this would be a good thing as this would favor the development of yet another style of animated mannerism. Animating by numbers is not better than painting by numbers. It become predictable quite fast.

Bottom line, I think this book is one of the must have books. I would have prefered if the book stayed more general but apart from that, it is still filled with very usefull information (take apart that eyeball rotation issue)