Computer animation

• 1. Introduction

  ›

  Computer animation is the process of digitally generating moving images, and it sits at the intersection of mathematics, art, and computing power. At its core, the technique works by displaying an image on screen and then replacing it rapidly with a slightly advanced version of the same scene. Do it fast enough, at around 24, 25, or 30 frames per second, and the human eye perceives fluid motion. Drop below 12 frames per second, and most people detect the jerkiness. That threshold, 12 frames per second, is the perceptual floor below which the illusion collapses.

  What makes computer animation distinct from its predecessors? How did it grow from experiments at a phone company lab in the 1960s into the dominant visual medium of today? And what actually happens inside the software when a character walks across the screen? Those are the questions this documentary will follow.

• 2. Bell Labs and Early Pioneers

  ›

  Bell Telephone Laboratories was where early digital computer animation took shape in the 1960s. The names attached to that work are Edward E. Zajac, Frank W. Sinden, Kenneth C. Knowlton, and A. Michael Noll. Separately, the Lawrence Livermore National Laboratory also pursued digital animation during the same period.

  In 1967, Charles Csuri and James Shaffer created a computer animation called "Hummingbird." A year later, in 1968, Nikolai Konstantinov used a BESM-4 computer to produce "Kitty," which depicted a cat moving around. In 1971, an animation called "Metadata" followed, showing various geometric shapes.

  The step that brought computer animation into cinemas came in 1976, with the film Futureworld. That film used 3D wire-frame imagery for a computer-animated hand and face, both created by University of Utah graduates Edwin Catmull and Fred Parke. The imagery had originally appeared in their 1972 student film, A Computer Animated Hand. Their previous experience came through a science-fiction production: the 1973 film Westworld, about a society in which robots live and work among humans, had preceded the sequel. Catmull's name would later become inseparable from Pixar.

• 3. How Motion Is Built

  ›

  Woody, the cowboy character in Toy Story, uses 712 animation variables, called Avars, with 212 of those controlling only his face. That number gives a sense of the underlying architecture of 3D character animation. Each Avar defines the position of one segment of a character's skeletal model, which is set up as a digital equivalent of a skeleton or stick figure. The computer uses these values to calculate the exact position and orientation of the character and then renders that into an image.

  Animators rarely set Avars for every single frame. Instead, they place values at strategic points in time and let the software interpolate between them. This process is called keyframing, and it has its roots in hand-drawn traditional animation. The software generates the in-between states automatically using splines that can follow Bezier curves to shape how motion flows between keys.

  A contrasting approach is motion capture. A real performer acts out a scene, and their movement is recorded using video cameras and markers. That recorded performance is then applied to the animated character. In the 2006 film Pirates of the Caribbean: Dead Man's Chest, Bill Nighy provided the performance for the character Davy Jones without appearing on screen. Motion capture preserves the subtleties of a particular actor, including posture and facial expression, while keyframe animation can produce movements that would be difficult or impossible to perform physically. As of 2007, productions were using either or both methods.

• 4. Modeling and Rigging

  ›

  3D models are built from geometrical vertices, faces, and edges arranged in a 3D coordinate system. Sculptors work from general forms toward specific details, much like working with real clay. Unless a model is a solid color, it is painted with textures to achieve realism.

  Rigging is the process of giving the virtual model controllers and handles so that animators can move it. Rhythm and Hues Studios spent two years creating the character Aslan for The Chronicles of Narnia: The Lion, the Witch and the Wardrobe. That model had roughly 1,851 controllers, with 742 of them in the face alone. For the 2005 remake of King Kong, actor Andy Serkis helped designers locate the gorilla in shots and contributed expressions that informed the creature's human-like qualities. Serkis had previously provided the voice and performance for Gollum in Peter Jackson's Lord of the Rings trilogy.

  For the 2004 film The Day After Tomorrow, designers had to simulate forces of extreme weather, drawing on video references and accurate meteorological facts. Each of these projects shows a different category of challenge: creature believability, emotional expression, and environmental simulation each require distinct workflows and specialized tools.

• 5. Facial Animation and the Uncanny Valley

  ›

  Realistic human facial features are considered among the most challenging and sought-after elements in computer-generated imagery. The first SIGGRAPH tutorials on the state of the art in facial animation, held in 1989 and 1990, brought together multiple research directions and sparked broader interest in the field. The Facial Action Coding System, which had been developed in 1976, became a popular foundation for many later systems. It organizes facial movement into 46 action units, covering expressions like a lip bite or a squint.

  By 2001, the MPEG-4 standard included 68 Face Animation Parameters covering the lips, jaws, and related features. More layered approaches followed. One approach uses the PAD emotional state model to assign emotions at a high level, then translates those down through a mid-level Partial Expression Parameters space, and finally into MPEG-4 facial parameters. Avatar: Way of Water offers a concrete illustration of where this work leads; WETA workshops designed digital muscles in their characters' faces with enough precision to match the emotional range of a human performer.

  Pushing human faces toward photorealism carries a risk. Audiences tend to respond with discomfort when a human replica looks and acts almost, but not quite, human. This effect is called the uncanny valley. Films including The Polar Express, Beowulf, and A Christmas Carol were criticized as disconcerting and creepy. The backlash shows that realism and believability are not always the same goal, and productions like Pixar's Turning Red took a different path, drawing influence from anime-style facial expressions instead.

• 6. Milestones in Film and Television

  ›

  CGI short films have been produced as independent animation since 1976. The first feature films to incorporate computer-generated imagery were Star Trek II: The Wrath of Khan and Tron, both released in 1982, along with the Japanese anime film Golgo 13: The Professional in 1983.

  The first full-length computer-animated television series was ReBoot, which debuted in September 1994. Its success followed VeggieTales, the first American fully 3D computer-animated series sold directly, made in 1993. ReBoot's distinctive premise placed its characters inside a computer. Transformers: Beast Wars followed in 1996 with a fully computer-generated style.

  Toy Story arrived in 1995 as the first feature-length computer-animated film, made by Disney and Pixar. It is considered a turning point for 3D animation. The character Woody in that film used 712 Avars to bring him to life.

  Decades later, Avatar in 2009 and The Jungle Book in 2016 used CGI for most of their runtime while keeping human actors. The Lion King in 2019 pushed further still, marketed as if it were a live-action film despite being computer-animated. Godzilla Minus One earned Toho Studios an Academy Award for its visual effects work achieved on a small budget relative to most comparable productions. The VFX team for Interstellar published a scientific paper about the mathematics used to create the black hole Gargantua.

• 7. The Infrastructure Behind the Screen

  ›

  Rendering a computer-animated feature film takes enormous computing resources. A workstation typically costs between $2,000 and $16,000, with more expensive machines rendering faster due to more advanced hardware. Silicon Graphics noted in 1989 that the animation industry's needs were consistently driving graphical innovations in workstations. Professional workstations use two to four processors and are specialized for rendering work.

  Productions group many such workstations together into what is called a render farm, networked so they operate as a single large computer. Even with a render farm, a complete computer-animated movie typically takes one to five years to produce, and rendering is only one part of that process.

  Software access has broadened the field. Programs like Blender allow people who cannot afford commercial-grade animation and rendering tools to work in a comparable way to professionals. On the web, animated GIF files represent the oldest backward-compatible format for browser-based animation. Flash animations dominated for decades before the web development community stopped supporting the Flash Player plugin. HTML5 technologies, including JavaScript and CSS animations, SVG animations, and APNG files, replaced those older formats and extended animation compatibility to mobile devices. APNG in particular offers multi-level transparency that GIF files cannot support. The rapid advancement of real-time rendering quality in games also led to a distinct art form: Machinima, where artists use game engines to render non-interactive films.