From Wireframes to Weta: The Evolution of CGI

When seeing Jurassic Park in 1993, audiences cried. Not because it was sad, but because it was impossible. For the first time, a dinosaur wasn't a puppet or a stop-motion clay model. It was breathing.

What is easy to forget is how little of that film is actually computer generated. By ILM's own account, the studio delivered fewer than 60 shots of fully CG dinosaurs. The revolution was not volume. It was credibility — and it arrived after twenty years of slow, unglamorous groundwork.

1973: The First Pixels (Westworld)

Decades before HBO's remake, the original Westworld movie used digital image processing to show the "Android Vision" of Yul Brynner's Gunslinger. The effect was produced by scanning film frames and rebuilding them as coarse colored blocks. It was not 3D and it was not modeled — it was a photograph taken apart by a computer, and it took hours of processing for every few seconds of screen time.

1976: The First 3D Model (Futureworld)

The sequel went further and became the first major feature to put 3D computer- generated imagery on screen. The animated hand was Ed Catmull's own left hand, digitized for his 1972 University of Utah short; the animated face came from Fred Parke's 1974 work. Catmull later co-founded Pixar, and that hand short was added to the National Film Registry in 2011.

This is the pattern that repeats through the whole history: the research comes first, often years earlier and in a university lab, and a film later borrows it.

1989-1991: The Liquid Era (The Abyss & T2)

James Cameron pushed ILM (Industrial Light & Magic) to create the water tentacle in The Abyss. It was the first realistic CG fluid rendering.

Two years later, they perfected it with the T-1000 in Terminator 2. The liquid metal cop wasn't just an effect; it was a character. This proved CGI could act.

1993: The Credibility Threshold (Jurassic Park)

Before Jurassic Park, no one had put living, breathing synthetic animals into a live-action film. Creature work meant animatronics and stop motion. The film crossed the line with a set of specific tools rather than one breakthrough.

ILM built Viewpaint so artists could paint texture directly onto a 3D model and see the result in context. Camera motion tracking removed the old rule that a background plate had to be locked off, so the camera could finally move in a shot containing a digital creature. And the Dinosaur Input Device, developed with Phil Tippett's team, let stop-motion animators pose a physical armature whose movement drove the CG model — a deliberate bridge between the craft that was ending and the one beginning.

The film was also the first major production to use the commercial Softimage animation package, and it won both the Academy Award and the BAFTA for its visual effects. Dennis Muren supervised. The lesson worth keeping is that the dinosaurs convinced people because of weight, light, and skin behavior, not because of resolution.

2009: Performance Capture (Avatar)

Avatar introduced "Performance Capture." Instead of just animating a character, head-mounted cameras recorded the actors' facial movement and mapped it onto 10-foot tall blue aliens. A virtual camera let Cameron walk the set and see a rough version of the digital world framed live. It moved CGI from "special effects" to "digital makeup."

Worth untangling: the motion-capture stage was called a volume, and today's LED-wall stages are also called volumes. They are different things. Motion capture records performance for later rendering. An LED volume displays a rendered environment on set so that it is photographed in-camera, lighting the actors and reflecting off real surfaces. That in-camera lighting is the actual advantage; it solves the reflection and eyeline problems that green screen never fully solved.

What a CG Shot Is Actually Made Of

"CGI" describes an outcome, not a job. A single creature shot passes through several distinct crafts, and a failure in any one of them is what audiences label as bad CGI.

Modeling builds the geometry. Texturing and look development decide how the surface responds to light — whether skin scatters, whether metal is scratched, whether anything is subtly imperfect. Rigging creates the skeleton and controls an animator can move. Animation supplies the performance and, above all, the weight. Lighting matches the digital element to the plate's real light sources. Rendering computes the result. Compositing then integrates it with the photographed frame, adding the grain, lens defects, and atmosphere that make two different images look like one.

This is why complaints about a film "having too much CGI" can miss the production problem. Tight schedules and late design changes can leave less time for lighting and compositing, regardless of how detailed the model is.

The Invisible CGI of Today

You think you hate CGI, but you only notice the bad stuff. David Fincher's Mindhunter and Parasite are full of CGI—changing backgrounds, adding blood, removing reflections.

The invisible category is now most of the work: removing a modern car from a period street, extending a set two floors higher, replacing a gray sky, erasing a safety wire, fixing a continuity error in a costume, or adding breath in a cold scene shot in summer. None of it announces itself, and none of it is what people picture when they argue about digital effects.

The best Visual Effects are the ones you never know existed.