The Problem Nobody Had Solved Before: Animating Liquid Metal
When James Cameron described the T-1000 to ILM’s visual effects supervisor Dennis Muren, he was describing something that had never been rendered before: a humanoid made of liquid metal capable of morphing, splitting, and reassembling in real time. Muren recognised immediately that nothing in ILM’s existing software library could handle it. The problem was not a matter of pushing familiar tools harder — the tools simply did not exist.
What followed was less a visual effects production than a research and development programme operating under blockbuster deadlines. ILM’s computer graphics department, based in San Rafael and still remarkably small by any modern standard, had to architect new software from the ground up before a single frame of the T-1000 could be rendered. Tools like Make Sticky and Body Sock — names that sound almost casual — were purpose-built solutions to specific, previously unsolved problems in CGI character animation: how to keep a morphing surface coherent across frames, and how to wrap a simulated skin convincingly over a transforming geometry.
The scale of the CGI work in T2 is routinely misread. The film contains a surprisingly small number of computer-generated shots. That low count is not a sign of modest ambition — it reflects how much foundational engineering each individual shot demanded. Every sequence required new pipeline decisions. There was no established workflow for simulating reflective liquid-metal surfaces, no shader library for chrome morphing, no precedent for blending a photorealistic digital double with practical photography at this level of complexity.
That context reframes what T2 actually represents in the history of digital visual effects. The film did not scale an existing process. It forced the invention of processes that the broader VFX industry would absorb, adapt, and build on for decades. The morphing algorithms, the surface-tracking approaches, the methods for rendering highly reflective organic geometry — these did not disappear when the production wrapped. They became the substrate on which later digital character work was constructed, from fluid simulation in feature animation to the muscle and skin systems used in contemporary digital humans.
‘Make Sticky’ and ‘Body Sock’: The Software Nobody Writes About
Ask any working VFX technical director to name the software tools that shaped modern character deformation pipelines, and you will almost certainly not hear “Make Sticky” or “Body Sock.” That obscurity is precisely the problem. Both tools were built by ILM’s computer graphics department in San Rafael during production on Terminator 2: Judgment Day, both solved problems that had no existing solution in production CG, and both seeded concepts that would become standard architecture in rigging and geometry systems used today.
Make Sticky addressed one of the most technically punishing aspects of animating the liquid-metal T-1000: keeping the character’s CG geometry coherently attached to a reference surface as the mesh transformed. Without it, the polygonal surface would drift, shear, or detach from its underlying anchor during a morph sequence, destroying the illusion of a continuous, cohesive material. Make Sticky enforced mesh continuity by binding geometry to a reference surface in a way that held through transformation — a problem of surface tracking and deformation constraint that animators and TDs working in the early 1990s had no off-the-shelf tool to solve. The software essentially invented the concept of sticky surface binding in a production pipeline context.
Body Sock operated one level up from that foundation. Rather than constraining geometry to a surface, it functioned as a deformation wrapper — a system that let animators drive complex, large-scale changes across a character model without manually keyframing individual geometry points. The computational cost and time involved in hand-animating thousands of vertices across a T-1000 transformation shot would have made the work impossible to complete on schedule. Body Sock abstracted that control, allowing high-level animation inputs to propagate through the mesh automatically. The conceptual architecture of that approach — wrapping a high-resolution target inside a lower-resolution deformation cage — is directly recognizable in the lattice deformers, wrap deformers, and cage-based rigging systems that populate every major 3D package used in production today.
Neither tool was documented formally. Neither was commercialized. They were built fast, used on one film, and left behind in ILM’s internal codebase. The ideas, however, did not stay there.
How ILM Organised an Invention Factory Under Production Pressure
Dennis Muren ran ILM’s visual effects operation on T2 less like a traditional production department and more like a skunkworks lab with a release date. His CG group in San Rafael was small — genuinely small, by any standard — yet he gave individual engineers and technical directors unusual freedom to prototype solutions without requiring proof of concept before work began. That management philosophy rarely gets named as a factor in the film’s success, but it was structural. Without it, tools like Make Sticky and Body Sock would never have existed in time to matter.
The process those TDs actually lived through was prototype, fail, rebuild, and ship — all within a single production cycle. An engineer would identify a problem the liquid metal terminator shots demanded, write a first-pass tool, watch it break against real animation and lighting conditions, and iterate. There was no separate R&D phase insulated from production pressure. The research was the production. Some tools went from initial concept to use in final-quality renders within the same compressed timeline that governed every other department on the film.
That model of embedded innovation — where experimental software development runs concurrently with live shot production rather than preceding it — is precisely what studios like Pixar and Weta Digital later institutionalised as deliberate policy. At ILM in 1991, it was a necessity driven by schedule. The oral histories from the artists who built those CG morphing and surface-deformation systems make clear that no off-the-shelf software package could handle what the T-1000 required. Every significant tool in the pipeline was written from scratch, tested on actual Terminator 2 sequences, and either proved itself or was replaced fast.
Muren’s latitude gave technical artists ownership over their solutions. That ownership accelerated decision-making in ways that top-down engineering management could not. The engineers knew what the shots needed because they were embedded in the shot-making process, not consulting on it from the outside. The result was a VFX pipeline methodology — iterative, production-integrated, creatively empowered — that predates the formal innovation frameworks modern studios now build into their studio infrastructure by design.
What Most Coverage Gets Wrong: Spectacle Over Engineering
Thirty years of retrospectives on Terminator 2: Judgment Day follow the same script. Writers reach for the chrome sheen of the T-1000, the floor-tile morphing sequence, the moment Robert Patrick’s face reassembles from a pool of liquid metal. Those images are genuinely extraordinary. They are also the wrong place to stop looking.
What mainstream film journalism consistently omits is the software architecture that generated those images in the first place. ILM’s computer graphics department in San Rafael — a unit that was, by any measure, astonishingly small for the scale of the problem it was solving — did not open an existing toolbox to build the T-1000. Tools like Make Sticky and Body Sock did not exist before production began. The team wrote them specifically because the creative problems demanded solutions that commercial software in 1991 simply could not provide. That is a software story, an engineering story, a research-and-development story. Film criticism has spent three decades treating it as a cinematography story.
The distinction carries real consequences. The development methodology ILM established on T2 — building production tools in-house, on-demand, as creative problems surface — is now standard practice at every major visual effects studio. Weta Digital, DNEG, Framestore, and ILM itself operate on exactly this principle. When a pipeline breaks or a shot requires a capability that no commercial package offers, technical directors write the tool. That workflow did not emerge from nowhere. It was pressure-tested and institutionalized during the making of Terminator 2.
The engineers and technical directors who built that toolkit — whose names do not appear alongside James Cameron’s or Dennis Muren’s in most anniversary pieces — built something more durable than any single visual effect. They built a model for how VFX production operates. Focusing exclusively on the spectacle of the liquid-metal Terminator while ignoring the CG pipeline innovations and proprietary software development behind it is not just an oversight. It is a systematic misreading of where the film’s actual legacy lives.
The Ripple Effect: T2’s Toolkit in the Context of Modern VFX
Body Sock’s logic — wrapping a control mesh around a character to drive surface deformation — is the direct ancestor of the blend-shape and cage-deformer systems that became standard across every major 3D package through the 1990s and 2000s. When Weta Digital built the digital creatures pipeline for The Lord of the Rings trilogy between 1999 and 2003, the deformation rigs controlling Gollum’s skin drew on exactly this principle: a secondary mesh influencing a hero surface, with simulation layered on top. James Cameron’s own Avatar (2009) pushed that inheritance further still, using simulation-driven muscle and fatty-tissue systems that are recognizable descendants of the same core problem ILM’s engineers were solving in San Rafael in 1990.
That lineage is rarely documented with any precision, because the engineers who wrote the original code were not the people being interviewed on press junkets. Oral history work — going directly to the programmers and technical directors who built tools like Make Sticky and Body Sock — surfaces institutional knowledge that lives nowhere else. When those practitioners retire or die, the knowledge goes with them. The VFX industry has no equivalent of the source-code archives or oral history projects that exist in, say, the games industry or academic computer science. Foundational decisions made at ILM in the early 1990s, decisions that shaped how every subsequent character effects pipeline was architected, exist primarily in the memories of a shrinking group of people now in their 50s, 60s, and 70s.
In 2024, that warning is immediately relevant. Generative AI tools are being folded into VFX production workflows right now, at speed, by small teams working under deadline pressure — exactly the conditions under which T2’s toolkit was invented. The practitioners building these new pipelines are moving faster than any journalist or industry historian is documenting them. T2 demonstrates the cost of that gap: it took more than 25 years after the film’s release for anyone to conduct a systematic technical oral history of how its effects were actually made. The industry cannot afford another 25-year delay before it understands what is being built today.
