Importing Traditional Animation Techniques into CGI

[1]

Shrek 2 hopes to fracture the fairy tale just like its predecessor. Photo courtesy of DreamWorks Pictures TM & © 2003 DreamWorks LLC. [1]

Computer generated imagery has become a daunting force in the world of animation. Certainly, the trend in general feature film production domestically favors semi-photorealistic computer animation. But while it is relatively natural to the medium to produce solid renditions of the world, the computer tends to lack the energy, emotion, and flexibility that traditional techniques enjoy more easily.

We’ll address innovations from the production and academic worlds of CG animation that draw directly from the lessons and aesthetics learned by traditional animators years ago. From animation studios such as DreamWorks to academic graphics research labs, we’ll explore techniques in motion, rendering and simulation that aspire to give computer animation the artistic control that 2D animation enjoys.

For an overview of traditional animation techniques, the reader is encouraged to reference Disney Animation: The Illusion of Life [2] by Frank Thomas and Ollie Johnston, as well as The Animation Book [3] by Kit Laybourne.

Motion Control: The Principles of Animation
Animators have been fighting to achieve more emotive results in computer animation almost since the first digital animators. Generally, it is accepted that the computer doesn’t change the fundamental principles of animation as addressed by the original Disney artists. It only transports those rules to a new world (see John Lasseter’s 1987 SIGGRAPH Presentation, “Principles of Traditional Animation Applied to 3D Computer Animation”).¹

The first step seems to always come down to the proper training. As John Canemaker [4], animation historian and director of animation studies at NYU’s Tisch School of the Arts, observes, “Animation still relies on the principles of motion. [The computer] is a tool as much as a pencil is.”

[5]

At NYU’s animation studies program, John Canemaker makes sure that students are trained in traditional and computer techniques. [5]

Canemaker adds that NYU has a very eclectic program, combining and intermingleing traditional and computer animation. “We feel we have to give them all those things to make them truly prepared. The more they know of all the possibilities, the better the production is going to be.”

This commitment to drawing upon the successful principles of motion in traditional animation echoes throughout many of the studios. AWN’s recent coverage of Glen Keane’s new Disney project, Rapunzel Unbraided, reports that the 2D veteran’s willingness to embrace the new medium allowed Keane and the other 3D artists to share ideas and try new perspectives.

In addition, Sony Pictures Animation evp, Sandra Rabins, told VFXWorld that the upcoming Open Season “is going to have more squash-and-stretch animation than we’ve seen or at least than [evp] Penny [Finkleman Cox] and I have been affiliated with...”

Similarly, director Vicky Jenson has stated that, with DreamWorks’ Shark Tale, “We want to go back and play with animation and do what it does best, which is the squash-and-stretch and the playfulness in the exaggerations you can do in animation.” This point has been underscored by DreamWorks partner Jeffrey Katzenberg in referring also to Shrek 2, Madagascar and Over the Hedge.

With a direction set, it comes down to an implementation problem. DreamWorks’ head of artistic development, Frank Gladstone [6], shares his approach at solving this problem. “To me, it’s almost a philosophical question,” says Gladstone. “You can do a lot in computer animation that can achieve the same results or those like 2D. But I don’t think in general training that those things are stressed as much as they could be. I wouldn’t develop a technology. I would develop the skills of the artists. That ability to nuance a performance rather than just [going] through the motions.

“Animation is the ability to exaggerate, to caricature life. It doesn’t mirror it. Technically speaking, we’re still not at that point when you always, easily get the performance. I think you can get the motion, but not the performance. It’s easy in traditional animation to move an eyebrow for a subtle expression. In CG, you have to fight a little bit to do that.”

Fight indeed. The benefit of being able to create reusable characters comes at a high cost of character preparation. The freedom of expression requires enormous control. For computer animation, every new control for a character requires the precious time of an articulator. The collection of controls for a character can quickly become overwhelming. For demands of exaggeration that fall beyond what can be achieved with current techniques, technology must step in.

A variety of point deformation techniques already exist to assist the character articulation artist in his task, including cluster transformations, lattice deformations, interpolated geometry, etc. While assembling these techniques into a structured control framework for production may give animators ample control, depending on the context of the application, such tools may need to be further augmented. More advances techniques of volume preservation and surface relaxation procedures may be required to blend such deformations into a more photorealistic character and maintain the realistic requirements of the production.

And while these technologies address some of the issues, other technology still needs to be developed in order to give animators the flexibility of a pencil.

For DreamWorks’ Frank Gladstone (left), change is good. Michael Fu (right) and Fluid are developing Putty, a technology that will help animators move characters’ facial features.

Michael Fu comments on a piece of technology being developed at Fluid called Putty. “Putty is designed to address some of the flexibility traditional illustrators have when constructing an appealing character design,” he comments. “The goal of Putty is to allow the animator to cheat a model as an illustrator in a two dimensional world would cheat surface features to give a pose a stronger silhouette.”

Putty allows the animator to easily move features such as an eye or nose or mouth along the surface of an object while preserving the general volume of the character. “We initially developed it for a character that was very much the equivalent of a bag of Jell-O. Because of the nature of the character, the character designer took liberties in facial expressions that would have normally been very time consuming to rig. Developing putty allows the features swim around the body to more easily hit those designs.”

Controlling Nature: The Sorcerer’s Apprentice
Almost everyone is familiar with Disney’s Fantasia and that marvelous scene in the Sorcerer’s Apprentice when Mickey Mouse commands the forces of nature. Or attempts to.

The traditional effects animator was tasked with creating the hyper-real — a spectacle that is more entertaining than reality. It was a laborious task that required the meticulous skill of representing natural phenomena that at the same time were both qualitatively realistic yet fundamentally narrative and thus “scripted.” As with the principles of animation, the meticulousness of the task has not changed. Nor have the goals. What has changed, however, is the technology.

While it is possible to hand-animate some of these effects and achieve total control (the rigid body dynamics of a car crash), more and more these effects are being determined by complex computational simulations. While this has the benefit of relying on the computer for what it does best — crunch insanely massive sets of data and equations — computational simulations are inherently difficult to control. Artists require high level, simple abstractions to control these simulations. And it looks like they are starting to get them. Professor Zoran Popovic of the University of Washington has been working on this problem for quite a while. He observes, “What I think we’re starting to do is sort of model all natural phenomenon in a way that you can control it. That’s the most fundamental thing, especially from the perspective of graphics animation.”

[8]

Zoran Popovic, Antoine Margerin McNamara and Adrien Treuille of University of Washington use animation to control gas images. © 2003 UW Graphics and Imaging Laboratory. [8]

In “Interactive Manipulation of Rigid Body Simulations” (SIGGRAPH Proceedings 2000)², a work led by Jovan Popovic with the assistance of Zoran Popovic and others, they propose a technique for introducing keyframe control for active rigid body simulation participants. The artist merely specifies the translations and rotations required for a ball or a flying car fender to hit specific goals in the animation. The system then inversely computes the initial and subsequent energies required to hit the artistic demands of the simulation while maintaining the quality of a natural, physical simulation.

These principles have been further extended to more complex systems of simulation such as gases, whose natural motion depends on solving complex systems of partial differential equations. To solve this problem and achieve the hyper-real performances so successful in traditional animation, we turn again to the University of Washington. In 2003, Zoran Popovic, et al. describe and demonstrate a technique in “Keyframe Control of Smoke Simulations” (SIGGRAPH Proceedings 2003)³ that gives an animator precise control over the visual direction of a body of gas by sketching shapes. The animator can precisely define the targets a body of gas must hit during a physical simulation.

Professor Popovic comments on the future direction of this work. “What we’re doing now is taking a look at incompressible fluids. We’re looking at water as well as more viscous types of things. The difference between smoke and water is the boundary and what is particularly different is that you can’t compress it like you can compress gas. Water has to preserve volume. That constraint turns out to be very hard to maintain so that volume doesn’t get lost in the process of simulation.”

Creating such high-level controls for computational simulations is challenging. “I think you end up designing different simulations such that you can control them, “ adds Professor Popovic. “In fact, what we’re finding is that we’re designing completely different simulators for water because the current ones, while okay for simulation, are not as powerful as they could be for control.”

[11]

Adam Finklestein leads a group project on NPR research. Image of a 3D still life from WYSIWYG NPR courtesy of Professor Adam Finkelstein of Princeton University. [11]

The Trompe L’Oeil Effect: Fighting Three Dimensions with NPR
French art critics in the 19th century coined the term trompe l’oeil for imagery that is so real that it fools the eye into believing that it is in fact a real object that can be touched.

Without a doubt, much of the fanfare for computer animation has come from photorealistic rendering techniques that give people this sensory experience. As Gladstone observes, “The lighting and texturing of things enamors people greatly. Those areas are astounding and only keep getting more polished all the time. That’s the strongest point.”

But a photorealistic effect is not always desired. Professor Canemaker comments, “In CG there is a tendency toward realism. It’s easy to go more real with things instead of styling it and making it more [organic]… You have to work really hard against a trompe l’oeil effect.”

With NPR, animators can combine the best of 3D, as in South Park, and traditional animation, as seen in A Charlie Brown Christmas. South Park © Comedy Central; A Charlie Brown Christmas courtesy of ABC / United Features Syndicate.

Non-photorealistic rendering research (NPR) offers a world of possibilities to those who wish to combine traditional techniques with the underlying technology of 3D computer animation. Audiences love Comedy Central’s South Park. Everyone adores the Peanuts television specials of yesteryear. And they appreciate the unique visual styles that each offers.

In the traditional world, the production intricacies of each style forced artists into disciplined containers. NPR research offers the hope that the same artists trained in traditional computer animation will be able to deliver a wide variety of final looks, whether that be traditional cel animation, sand painting or stop motion.

The simplest of these methods is the cel-shaded look. Objects are rendered with traditional systems, but employ shaders to generate a look akin to the finite shaded tones of a cel-painted character. Such techniques were formerly still traditionally rendered. The new frontier in graphics is happening in the interactive space.

Santa Clara-based Nvidia Corp. and Toronto-based ATI are both manufacturers of high-end graphics chips and are leading the development in this area. By extending and accelerating the shading capabilities of their graphics hardware pipelines, they have allowed programmers to develop realtime shaders. Although photorealistic effects tend to be more complicated, possibly requiring multipass algorithms, cel-shaded techniques are computationally simple enough to pull off in realtime.

But cel shading can only go so far. In particular, the rendering of the surface is confined to boundaries of the surface silhouette. To truly stylize an object, it is necessary to take control of the contour and silhouette lines that traditionally define an illustration.

It is here that researchers at Princeton are working on several projects that move beyond the simple arena of cel shading. To date, the collaboration led by Professor Adam Finkelstein has yielded many useful techniques for automatically determining an objects’ lines, giving them user-defined, artistic properties that can be automatically and easily applied to the entire model. In addition, their research allows the artist to draw feature elements on an object much the way an illustrator would provide hatching to convey shadow properties.

Examples of the group’s work can be found in “WYSIWYG NPR” (SIGGRAPH Proceedings 2002), “Coherent Stylized Silhouettes” (SIGGRAPH Proceedings 2003) and “Suggestive Contours for Conveying Shape” (SIGGRAPH Proceedings 2003)⁴. In addition, software is available for download from their Web site (jot.cs.princeton.edu).

The Future
So what will the future present? Surely, refinements and advances in technology will allow computer animation to produce ever more amazing images. Certainly, classical training should continue to be stressed.

If the past 10 years is any indication, the level of sophistication that the digital artist brings to the table in terms of a range of techniques will get ever more rich. With each advancement, the techniques available to artists from traditional animation will soon become available to computer-animated artists.

And when that occurs, one hopes to see these techniques explode as computer animators move beyond photorealistic rendering, much the same way traditional artists move beyond their realist periods. After all, it was only then that the Chagalls, Picassos and Miros of the world defined themselves.

DreamWork’s Gladstone offers some perspective for those of us shy of change. “For myself, seeing the changes in 30 years-plus in the business is scary when I first see them and then absolutely wonderful with I see them integrated into film. The business of creating these images is only getting more engaging. I do see a room for all of the techniques.”

Seth Piezas is president of Fluid, a production management and technology company specializing in high-end, computer-animated feature film. Seth recently departed Pixar Animation Studios where he was a member of their production optimization team. His credits include Monsters, Inc. and Toy Story 2.

Contributing authors Michael King and Michael Fu are partners in Fluid. They also recently hail from Pixar Animation Studios.