Skip to content

The CGI Revolution

Abstract

Computer Generated Imagery began as an academic experiment and became the dominant visual language of Hollywood. Ed Catmull, Alvy Ray Smith, and James Cameron pushed forward a technology that systematically replaced miniature models, matte paintings, and stunt performers, until a film about digital humans failed at an invisible psychological threshold and exposed the limits of the technique.

The Academic Roots: Catmull and the Polygon Problem

Ed Catmull started graduate work in computer graphics in 1970 at the University of Utah, one of the few places in the world where computer graphics counted as a serious research field. His dissertation contained two innovations that would shape the entire industry: Z-buffering (an algorithm for correctly rendering the depth of overlapping objects) and texture mapping (projecting two-dimensional images onto three-dimensional surfaces).

The fundamental problem of early computer graphics was the gap between mathematical description and visual realism. Polygons (the basic building blocks of three-dimensional computer models) produced faceted, angular surfaces. Organic shapes, curved surfaces, the flowing contours of a human face: none of it could be represented well with coarse polygon meshes.

Catmull, with Jim Clark, published subdivision surfaces for this in 1978, an algorithm that iteratively splits coarse polygon meshes into ever finer, smoother curves. Repeated subdivision turns a cube into a sphere. A rough shape becomes organic complexity.

Alvy Ray Smith, co-founder of what would become Pixar, brought the complementary expertise: color theory, alpha compositing (the ability to combine computer images with film footage), and an understanding of how digital and analog image worlds could grow together.

The field’s academic infrastructure emerged in parallel: Andries van Dam (Brown University, in 1967 co-founder of ACM SICGRAPH, the predecessor of SIGGRAPH; see also Ted Nelson and Hypertext) wrote with James Foley the standard textbook Fundamentals of Interactive Computer Graphics (1982), which from its second edition in 1990 appeared as Computer Graphics: Principles and Practice and is called “the bible of computer graphics” for its authority and scope.

The Commercial Branch: Triple-I and the Cray Years

Academic graphics reached the screen through a small commercial industry that no longer exists. Information International, Inc. (Triple-I), a Los Angeles maker of precision film recorders and scanners, ran a motion-picture group under Gary Demos and John Whitney Jr. that put digital image processing into Westworld (1973) and three-dimensional computer graphics into Futureworld (1976), then produced much of the computer animation in Disney’s TRON (1982) alongside MAGI.

Demos and Whitney Jr. left during TRON to found Digital Productions, which bought an early Cray X-MP in October 1983 (see Seymour Cray and Supercomputing) and used it to render about 27 minutes of imagery for The Last Starfighter (1984), where the spaceships are geometry rather than models. The industry then destroyed itself in a single transaction: Omnibus Computer Graphics bought Digital Productions and Robert Abel and Associates in 1986, ran out of money, and closed all three studios on 13 April 1987. When Pixar and ILM took over the field in the following years, they inherited a market with no incumbents.

Lucasfilm and the Birth of Pixar

In 1979, George Lucas brought Catmull and Smith to San Francisco to build a computer graphics division at Lucasfilm. The goal was pragmatic: Lucas wanted to produce special effects more efficiently. What emerged was something more fundamental.

The Lucasfilm Computer Division produced, in 1982, the first fully computer-animated film sequence with photorealistic ambitions: the Genesis effect in Star Trek II: The Wrath of Khan, a 60-second sequence showing a barren planetary surface transformed by a chain reaction into a living biosphere. Technically it was a milestone. Artistically it was a demonstration: computers could create imagery no model maker could reproduce.

In 1986, Steve Jobs bought the computer graphics division for 5 million dollars from Lucas, whose divorce had left Lucasfilm short of cash. The Lucasfilm division became Pixar. Jobs invested another 5 million as working capital. He expected to sell hardware systems; Catmull and Smith saw in Pixar a film studio.

Two years later, in 1988, Pixar published the RenderMan specification, a description language for three-dimensional scenes that worked like a universal translator between 3D model and rendered image. RenderMan became the industry standard for photorealistic rendering: over the decades, a large share of the films that won the visual effects Oscar were rendered with it.

ILM and the Digital Stunt: Terminator 2

Industrial Light & Magic (ILM), Lucas’ special effects studio founded for Star Wars, became the second center of CGI development. While Pixar worked on pure computer graphics, ILM focused on integrating digital elements into filmed footage.

The breakthrough came in 1991. James Cameron produced Terminator 2: Judgment Day for roughly 70 million dollars of production cost, 94 to 102 million including marketing, at the time the most expensive film ever made. His central visual concept was the T-1000: a Terminator made of liquid metal that could morph into any shape. That was physically impossible to film.

ILM had already built a convincing digital creature for Cameron, the water pseudopod in The Abyss (1989), which won that year’s visual effects Oscar. The T-1000 was the step from a creature to a character. Played by Robert Patrick and transformed through CGI, it walked through solid metal, flowed through prison bars, reassembled its face from metal shards. Roughly five minutes of screen time across 42 or 43 CGI shots cost about 5 million dollars, out of 15 to 17 million for effects overall; a team of up to 35 people at ILM produced it, and rendering fifteen seconds of footage could take ten days. By the standards of the day, that was a tiny amount of screen time for enormous effort. But the effect was undeniable.

T2 won the Oscar for Best Visual Effects. More important: it proved that digital characters could exist credibly next to real actors. The foundation for what would follow two years later was laid.

Jurassic Park: The Turning Point

In 1993, one film changed the industry permanently. Steven Spielberg produced Jurassic Park, the adaptation of Michael Crichton’s novel about a theme park with cloned dinosaurs. The production team originally planned to work with go-motion puppets and animatronic figures, the state of the art for life-sized creatures.

Phil Tippett, one of Hollywood’s leading effects artists and a go-motion pioneer, had already worked for months on dinosaur sequences when ILM staff showed Spielberg a computer-generated test: a galloping herd of Gallimimus. Spielberg was convinced on the spot. Tippett is said to have told Spielberg: “I think I’m extinct.” Spielberg wrote the line into the film.

Jurassic Park was the first film in which computer-generated creatures appeared in natural outdoor light, next to real actors, with photorealistic results. ILM had to develop entirely new rendering techniques: muscle deformations that responded plausibly to movement; skin with texture and reflective properties; lighting models that integrated digital objects into real environments.

The result convinced audiences and critics completely. The dinosaurs of Jurassic Park looked more credible than any model, puppet, or costume ever could, because computers allowed unlimited polygons, unlimited complexity of movement, unlimited fineness of texture detail.

Info

The uncanny valley: the psychological limit of computer graphics

The Japanese roboticist Masahiro Mori described a phenomenon in 1970 that he called bukimi no tani (the “uncanny valley”). His thesis: the more human-like a figure becomes, the more sympathetic we find it, up to a point. Just short of complete human likeness, the sympathy curve plunges into a deep valley of unease, strangeness, sometimes disgust. Only when a figure becomes practically indistinguishable from a real human does sympathy rise back to normal levels.

Mori illustrated this with robots, but the concept precisely describes a problem CGI productions have experienced systematically. Stylized, clearly non-realistic characters (Pixar’s characters, Shrek) generate sympathy. Photorealistic animals and creatures (Jurassic Park, The Jungle Book) work because no human baseline for comparison exists. Digital humans who look almost realistic, however, trigger instinctive unease, an evolutionarily old signal pointing to sickness, deception, or decay.

For CGI productions, the uncanny valley is not a philosophical question but an engineering barrier: investing in more realism makes the character worse, until it crosses the threshold of complete conviction, which for human characters remains technically and economically extremely difficult to this day.

Toy Story: The Fully Digital World

While ILM developed CGI as a complement to filmed material, Pixar pursued a more radical goal: a fully digitally produced feature film without a single piece of film footage.

Toy Story (November 1995, directed by John Lasseter) was the result of four years of development and a bet that was existential for Pixar. Disney had signed a contract in 1991: three feature films, Disney markets and distributes, Pixar produces. The budget for Toy Story was 30 million dollars, a fraction of an average Disney hand-drawn animation production.

The technical challenges were considerable. Human-like characters (Woody, Buzz Lightyear as toys) had to be animated convincingly without organic softness; they were deliberately conceived as toys, which permitted stiffness and plasticity. Hair and clothing were simplified. Backgrounds were geometric. Lasseter and his team built in artificial constraints that simultaneously solved the production problem and looked aesthetically coherent.

Toy Story has taken just over 401 million dollars worldwide across its 1995 release and later reissues, on a 30 million production budget. The film earned John Lasseter a Special Achievement Award from the Academy of Motion Picture Arts and Sciences. An entire industry began rethinking its production pipelines.

Toy Story was followed by A Bug’s Life (1998), Toy Story 2 (1999), and Monsters, Inc. (2001), each film technically more demanding than the last. Each film proved that the fully digital format was economically and artistically competitive.

Getting Film In and Out

None of the compositing work was possible without a step that never appears in a making-of documentary: reading a live-action negative into a computer at full quality and writing the treated result back onto film so that the doctored shots and the untouched ones look identical in the same reel. A scanner that loses shadow detail or a recorder that shifts color makes every effect visible as an effect, and no amount of rendering fixes it.

The Academy recognized the people who solved it with a Scientific and Engineering Award for 1994, shared across three rival organizations: Gary Demos and Dan Cameron of Information International, David DiFrancesco and Gary Starkweather of Pixar, and Scott Squires of ILM, “for pioneering work in the field of film input scanning.” Squires had come to ILM from Dream Quest Images in 1985 and stayed twenty years as a visual effects supervisor, with Oscar nominations for The Mask, Dragonheart and Star Wars: The Phantom Menace; he pushed the scanning work from the production side. DiFrancesco built the laser film recording that closed the loop in the other direction, and took his own Academy Technical Achievement Award for it in 1999.

Once the round trip was reliable, it stopped being an effects tool and became the standard way films were finished. The digital intermediate, scanning an entire feature, grading it digitally and recording it back out, replaced photochemical color timing in the early 2000s. O Brother, Where Art Thou? (2000) was the first first-run Hollywood feature put through the process end to end: Cinesite scanned it at 2K on a Spirit DataCine, Roger Deakins spent eleven weeks pulling the greens toward burnt yellow, and a Kodak Lightning II recorder wrote it back to film.

The Systematic Replacement of Traditional Techniques

The 1990s and early 2000s saw a systematic retreat of traditional film craft before the expansion of CGI:

Miniature models (for decades the standard for spaceships, explosions, cityscapes) were increasingly replaced by digital models. Physical models could exist only once, could be filmed at only one scale, broke in accidents. Digital models existed indefinitely, at any scale, with unlimited reproducibility.

Matte paintings (hand-painted backdrops that placed studio productions in exotic settings) gave way to digital environments. Photoshop and 3D background rendering enabled backdrops that moved with the camera and in which light changed dynamically.

Stunt performers were not replaced, but their roles were redefined. Dangerous falls, explosions, crowd scenes: CGI took over what was practically infeasible or too risky. Wire removal (digitally erasing safety cables from shots) became standard postproduction.

Actors were augmented by digital doubles. The death of Oliver Reed during the production of Gladiator (2000) was bridged by digitally reconstructing his face on another actor’s body, a precedent with far-reaching ethical and contractual consequences.

Dead End: Final Fantasy: The Spirits Within

Warning

Final Fantasy: The Spirits Within: the uncanny valley as a 137-million-dollar lesson

In 2001, Hironobu Sakaguchi (creator of the Final Fantasy video game series) attempted something no feature film had dared: a fully computer-animated film with hyperrealistic human protagonists that an audience was supposed to accept as credible.

The result, produced by Square Pictures in Honolulu, cost 137 million dollars to make, a budget competing with the majors. The technical achievement was undeniable. The protagonist, Dr. Aki Ross, had 60,000 individually simulated hairs. Facial movements were transferred from real actors via motion capture. Skin surfaces showed pores and imperfections.

The audience reacted with cold distance. The characters looked almost human, and precisely that almost was the problem. Dr. Ross’ eyes moved in a way that felt mechanical. Smiles had a delay that a real brain registered as wrong without being able to name it. Skin reflected light incorrectly. Hair moved too uniformly.

Final Fantasy: The Spirits Within grossed 85.1 million dollars worldwide against a 137 million dollar production budget plus about 30 million that Columbia Pictures spent on marketing. Losses were estimated above 94 million dollars. Square Pictures closed in January 2002, and the loss delayed the Square-Enix merger, with Enix reluctant to combine with a company carrying that much damage. Sakaguchi resigned his executive position at Square in 2001, left the company in 2003, and founded the studio Mistwalker in 2004.

The film demonstrated what the uncanny valley means in practice: every dollar Square invested in more realism brought the characters closer to the valley without crossing it. The goal (complete conviction) lay beyond the technical and economic possibilities of the time. A strategy of deliberate stylization (as at Pixar) would have sidestepped the problem. Sakaguchi’s team chose the valley instead.

Legacy

The CGI revolution changed not only film production but the whole of visual culture. Video games, advertising, architectural visualization, medical imaging, product design: wherever three-dimensional visualization played a role, the tools and techniques of the film world were adapted.

By 2005, Hollywood had converted practically all major productions to hybrid or fully digital effects. The training pipelines for CGI artists, the software industry (Autodesk Maya, Houdini, Nuke), the render farm infrastructure: all of it grew into an industry of its own.

Ed Catmull and John Lasseter led Pixar until Disney’s acquisition in 2006 for 7.4 billion dollars. Catmull became President of Walt Disney and Pixar Animation Studios. The academic experiment from the University of Utah had turned into one of the most valuable creative brands in history.

The open question (how convincing digital humans can ever be) remained. Performance capture, advanced by Robert Zemeckis (The Polar Express, 2004) and James Cameron (Avatar, 2009), partly bypassed the uncanny valley by transferring the real-time movements of real actors directly onto digital characters. Full conviction for human characters remained the industry’s unsolved problem.

πŸ“š Sources