Terminator 2 (1991) Visual Effects Oral History: How ILM Invented the Liquid Metal T‑1000

Terminator 2 (1991) Visual Effects Oral History: How ILM Invented the Liquid Metal T‑1000

TL;DR – Why Terminator 2 still matters

The 1991 film Terminator 2: Judgment Day forced Industrial Light & Magic (ILM) to invent an entire toolbox—Make Sticky, Body Sock, the poly‑alloy shader, MORF, and custom scripting pipelines—to render the liquid‑metal T‑1000. Those inventions solved problems that had never been tackled before and became the foundation for today’s character‑animation and compositing workflows.

1. The ILM team was tiny but ambitious

“We were a dozen‑odd people in CG and had to grow the department very quickly.” – George Joblove

  • In 1990 ILM’s computer‑graphics staff numbered 12‑15; by the end of the shoot it had swelled to ~300 across the studio.
  • The crew came from diverse backgrounds—Pixar, Alias, Pixar Image Computers, and even the original Lucasfilm graphics group—so they were both artists and early software engineers.
  • Dennis Muren’s confidence in the team’s ability to solve “impossible” problems set the tone: “If we couldn’t do it on The Abyss, we wouldn’t even try on T2.”

2. Hardware constraints drove software innovation

“A gigabyte of storage cost $9,000 in 1990.” – George Joblove

  • ILM relied on a network of SGI workstations (VGX‑340, VGX‑240) and a handful of large servers.
  • Rendering farms were built from command‑line shape‑animation tools and custom processor‑allocator (PA) GUIs to share CPUs across artists.
  • Disk swaps required physically changing platter drives in the basement, a process that could take minutes per change.

3. From The Abyss to Terminator 2: Re‑using and dissecting tools

“We took the pseudopod program from The Abyss and pulled it apart into separate utilities.” – John Schlag

  • The pseudopod (water‑creature) tool from The Abyss provided a single‑purpose pipeline for a creature. ILM modularized it into:
    • Make Sticky – a UV‑sticking system that kept texture coordinates attached to moving geometry.
    • Body Sock – an automated stitching algorithm that blended adjacent NURBS patches each frame, preventing gaps when the character’s limbs moved.
    • Z‑ripple – a procedural rippling utility for bullet‑hole healing, derived from the Abyss water‑ripple code.
  • These tools turned a single‑purpose program into a library of reusable components for the T‑1000’s many shape‑shifts.

4. The “poly‑alloy” shader: early programmable reflections

“We wrote a RenderMan shader that let TDs place multiple reflection planes and hit‑test them per frame.” – Alex Seiden

  • Real‑time ray‑tracing was far too slow; instead the team built a custom RenderMan shader that simulated chrome by projecting reflection planes onto the surface.
  • The shader supported multiple reflection maps, allowing the T‑1000 to reflect flames, fire, and environment geometry without full ray tracing.
  • Dennis Muren guided the look, insisting on a balance of diffuse shading and specular highlights to give the liquid metal a sense of mass (“the pewter look”).

5. Data capture and the lack of motion‑capture

“We painted a 4"×4" grid on Robert Patrick, shot him with two VistaVision cameras, and rotoscoped every frame.” – Steve ‘Spaz’ Williams

  • No modern motion‑capture existed; the team used hand‑digitized scans and dual‑camera rigs to capture body and facial motion.
  • Five T‑1000 stages (RP1‑RP5) were built from the same control‑vertex set, enabling model interpolation (what the film called “morph”) rather than true vertex blending.
  • The infamous limp from Patrick’s football injury was manually corrected in the rotoscoped data, illustrating the painstaking attention to detail.

6. Body Sock – stitching the liquid metal

“Body Sock was a ‘stretchy nylon fabric’ that automatically stitched all seams each frame.” – Eric Enderton

  • The T‑1000’s surface was composed of four‑sided B‑spline patches; as the skeleton moved, patches would separate or intersect.
  • Body Sock read a sock file describing which edges to stitch, then stitched every frame automatically, handling simple two‑surface seams and more complex multi‑surface corners.
  • This eliminated the need for artists to manually adjust UVs or re‑model each frame, a task that would now be a single click in modern NURBS or subdivision tools.

7. MORF – early shape‑interpolation UI

“MORF showed two windows (source and destination) and let us drag grid points to morph geometry.” – John Berton

  • MORF, originally built for Willow, was ported to SGI and used to morph the T‑1000’s geometry between stages.
  • Artists placed control‑grid points on source and target meshes; the tool performed bicubic spline interpolation, with visual feedback on overshoot and ringing.
  • The UI allowed per‑frame key‑framing of grid points, enabling the iconic “turn‑around” and “head‑through‑bars” transformations.

8. Make Sticky – early UV‑locking

“Make Sticky remembered texture coordinates at one frame and kept them stuck as the geometry moved.” – Steve ‘Spaz’ Williams

  • Traditional UV mapping attached textures to static vertices, causing sliding when the mesh deformed.
  • Make Sticky stored the UV coordinate for each point at a reference frame and re‑applied it each subsequent frame, effectively locking the texture to the moving surface.
  • This technique was crucial for the head‑through‑bars shot, where the T‑1000’s chrome needed to stay visually consistent while the geometry twisted.

9. Custom pipelines and compositing scripts

“Every compositing operation was a separate command‑line program that wrote to a shared virtual frame buffer.” – Doug Smythe

  • The team built a shell‑script‑driven compositing pipeline: load image → load matte → composite → write → repeat.
  • A virtual frame buffer allowed multiple processes to share image data without writing to disk, dramatically speeding up iterative work.
  • “Dave” the paint‑fixer would manually retouch frames in early Photoshop, a practice that earned the tongue‑in‑cheek nickname “model, animate, render, composite, Dave.”

10. The melting‑metal finale – fractals and last‑minute decisions

“We used random fractal displacement; Dennis wanted specific parts to dissolve faster, so we kept re‑seeding until it looked right.” – Michael Natkin

  • The final molten‑steel death sequence combined fractal‑generated displacement maps with hand‑tuned seeds to control melt speed.
  • Rendering was limited to ~1 K resolution; the final cut was a low‑res version that the studio shipped because of schedule pressure.
  • Josh Pines performed a scaling‑and‑printing trick to get the low‑res frames onto 35 mm release prints.

11. Legacy and cultural impact

“T2’s VFX still hold up; the tools we built became the basis for modern character pipelines.” – Geoff Campbell

  • The film won the Academy Award for Best Visual Effects and proved that CG characters could carry a narrative.
  • Many of the custom tools (Body Sock, Make Sticky, MORF) were later absorbed into commercial packages (Maya, Houdini, RenderMan) and are now taken for granted.
  • The oral history shows that software engineering, artistic intuition, and long hours were all essential; the team’s “DIY” ethos is echoed in today’s indie VFX houses.

12. Community reflections (Hacker News comments)

  • Users noted that the custom liquid‑metal bullet‑impact squibs remain among the best practical effects.
  • Several commenters highlighted the upcoming 4K/3D re‑release for the film’s 35th anniversary, confirming the lasting audience interest.
  • A few pointed out that Softimage was also used for some shots, expanding the toolset beyond Alias.
  • Others praised the documentary “Jurassic Punk” (2022) for covering Steve ‘Spaz’ Williams’ role on T2 and later on Jurassic Park.

Bottom line: Terminator 2 forced a small, resource‑starved ILM team to invent a suite of groundbreaking tools—Make Sticky, Body Sock, poly‑alloy shader, MORF, and custom compositing pipelines—that solved problems no one had faced before. Those inventions not only made the iconic liquid‑metal T‑1000 possible but also laid the groundwork for the modern VFX pipelines that power today’s blockbuster films.

Sources