T3: Creatively and LogisticallySpeaking
Terminator 3 Evolves Historic Effect
T3 visual effects supervisor Pablo Helman says that buildingupon Terminator 2’s legendary transformation effectrequired eight months of intensive R&D.
While 1991’s Terminator 2 is largely recognized as a seminalstep for the visual effects industry because of its then-groundbreakingmorph effect, it is largely forgotten that the movie featured fewerthan 50 digital effects shots total. Its long-awaited descendent,Terminator 3: Rise of the Machines, on the other hand, hasapproximately 650 digital effects shots, virtually all of them far morecomplex than anything seen in T2.
In particular, the shiny new T-X (Kristanna Loken) goes far beyondArnold Schwarzenegger’s original Terminator, or Robert Patrick’s T-1000from T2, from a visual effects point of view. But what’s thesignificance of the T-X liquefaction/transformation effect developed byILM, compared to T2‘s legendary morph?
“A lot of what we’ve done definitely contributes to theadvancement of our industry, but the difference is, it’s a differentindustry now than when ILM worked on T2,” responds PabloHelman, the film’s visual effects supervisor. “The kind of visualeffects audiences are used to, the kind of effects directors are askingfor, are all different — the demand now is for a wider variety ofeffects and more realism. In this movie, for instance, we use digitaldoubles, which is hardly new, but we get closer to them, to theirfaces, than we are used to doing. These are important developments, butthey are now expected by audiences. I periodically talk to [T2‘svisual effects supervisor] Dennis Muren, who works a couple officesover from me [effects supervisor for The Hulk — see theJune Millimeter]. He keeps telling me, ‘Don’t live in theshadow of T2; do your own project.’ That’s what we aredoing, and along the way, we are solving some extremely difficultproblems, but they are all related to this story.”
At the center of this effort is T-X’s transformation, somethingHelman insists they are not calling a morph. It is, however, in manyways a sophisticated descendent of T2‘s breakthrough effect,though achieved in an entirely different way. According to directorJonathan Mostow, the notion of allowing T-X to melt into other shapeswas simply a logical progression.
“What was important was creating a villain robot that would becompelling, both visually and viscerally,” says Mostow.“Obviously, we wanted something audiences hadn’t seen before, sothat, by definition, required new visual effects.
“The morphing liquid metal technology was something I wantedto keep — in some form — because I felt it afforded the T-Xthe useful ability to disguise herself. If [the previousTerminator film] showed they had a great technology in thefuture with the T-1000, why would they abandon it? At the same time, Ididn’t want to retread it. Instead, we figured out what the next stepin its evolutionary process might be, and then we worked with ILM torealize that visually,” he says.
Helman says that work first involved eight months of difficultR&D to come up with the effect. “The character has theabilities of the previous Terminators, including the liquid metalexterior and the armored endoskeleton and other things, but thequestion was how to have her transform from one thing to another. TheR&D effort came up with a fluid simulation approach —technology to control every single string of liquid.”
According to the film’s associate visual effects supervisor SamirHoon, existing fluid simulation solutions were developed for big wavesand splashes, like in The Perfect Storm. “Here, on theother hand, we needed to have flowing, liquid tendrils —localized areas. We needed controlled, gradual liquefaction of herexterior, slowly revealing the underlying machine in an acutelydirected manner.”
To achieve this, ILM’s team (led by Hoon, sequence supervisor NigelSumner, and software engineers Nick Rasmussen and Sebastian Marino)worked with their R&D consultant, assistant professor Ronald Fedkiwof Stanford University’s computer science department. Fedkiw and hisstudents had recently developed a fluid simulation approach known as aparticle level set (PLS) engine. Hoon describes it as “a methodthat uses a level-set, stored on a regular volume grid to represent thefluid/air interface. This interface is augmented with particles [usingMaya particles] on either side to help prevent volume loss as thelevel-set is advected through the fluid velocity field. The level-setis an implicit representation for a surface that is convenient forrepresenting surfaces that break apart and join together overtime.”
Hoon adds that the process can be painstaking but also moreaccurate, and therefore more realistic than other approaches.
“For a 3D fluid simulation, as the fluid is animated throughthe grid, numerical imprecision can result in a loss of fluidvolume,” he explains. “Typically, this error is a knownquantity, but is simply too small to represent given the fixedresolution of the grid being used for the simulation. For PLS fluid,when errors occur, spherical particles with a volume equivalent to thevolume lost from the level set are emitted from and are advected alongwith the fluid interface until the particles can be combined back intothe grid without loss.”
In other words, the approach permitted ILM to retain the volume offluid once it vacated a grid portion of the T-X computer model,allowing them to render it out with detailed, unique movement.
To achieve a sufficient level of control over the simulation, Hoonsays, “We broke the T-X’s body into smaller, multiple overlappingsimulation grids. The fluid moved identically in the overlapping areas,thereby allowing us to join them together after they were independentlysimulated. This permitted us to run high-resolution simulationsovernight that would have taken weeks to simulate on normal desktopmachines. We used Maya particles to control and direct the fluid bydispersing the particles over the outer surfaces of the T-X character,and these particles carried animation-dependent properties —position, velocity, viscosity, surface normals — that provided uswith a generic method for inputting control properties into the fluid.Exported particles were then used as controls for defining regions ofthe flow, with varying viscosity and areas where the fluid wasconstrained to track the particle motion. The texturing problem wasovercome by advecting these particles through the fluid, and then usingthem as texture coordinates for rendering. Geometry was used as fluiderasers, or emitters, to enhance the degree of control.”
T3: Creatively and Logistically Speaking
DP Don Burgess chose Kodak Super 35 with 200 ASA Tungsten stock forT3’s effects shots.
No matter the quality of Terminator 3, many people no doubtwill have a hard time forgetting that the film has a different directorand DP than the first two movies in the franchise. With Jonathan Mostow(U-571) stepping in for series creator James Cameron, andindustry veteran Don Burgess (Forrest Gump, Contact, Spider-Man)taking over for Adam Greenberg, fans are asking: Does this look andfeel like a Terminator movie?
“We have the star [Arnold Schwarzenegger] and a strong storyand great effects, but I view it like a legendary band on a new concerttour,” says Mostow. “Some people want to hear the oldclassics in exactly the same rendition they have heard a million times,and others want to hear updated versions. You can’t possibly makeeveryone happy, but the thing about this franchise is it’s not like theJames Bond films, which are set in their ways and have aformula. Here, the fun thing for me was deciding what should be part ofit and what should not be. What made the first two movies so incrediblefor me as a fan was the naturalistic shooting style and the highlyrealistic visual effects. We have all that — a similar style— but not necessarily the same look and color palette.”
Mostow and Burgess say this film has less of the cold blue palettethat the previous films featured. “I seem to gravitate towardmaking visceral films, so it felt natural to me to push T3toward a warmer temperature,” says Mostow.
The director says the shoot had “many layers of logisticalcomplications, more than I was used to, including the detailed use ofsophisticated computer-generated effects and animation.” He alsoadmits to a deep and fast learning curve in working with ILM on thosesequences, but he had a grizzled veteran by his side in Burgess.
“I learned about a whole new league of visual effects when Iworked with [Robert] Zemeckis on Forrest Gump, so I was wellprepared to work closely with ILM on the effects,” Burgess says.“Most of the visual effects plates had to be done first unitanyway because they involved principal actors. In truth, the technologyand its capabilities has changed greatly for what they can do with thedigital shots, but in terms of production, the approach hasn’t changedmuch from [the days of Forrest Gump]. It’s mainly about havingeverything properly storyboarded out and having sufficientconversations between all the key people — myself, the director,the visual effects supervisor, the production designer, and so on— about how to visualize and accomplish the shot so that theeffects people can composite it all together. We pushed the envelopequite a bit in terms of the nature of the effects, but the way we gotthere was the way we’ve always done it — with lots of hard,detailed work.”
Much of that hard work for Burgess revolved around the key cemeteryshootout sequence.
“That sequence was meant to be big, so I broke out 12 camerasand we staged it as one big master shot,” says Burgess. “Iused all those cameras to cover the wide, medium, and tight approaches,all at the same time. That gave us options to do lots of cutting toimprove pace and tension.”
Burgess shot the movie on Kodak Super 35 with 200 ASA Tungsten stockfor all the effects shots. “The 200 ASA Tungsten stock gave usresults like medium grain, features that would hold up for the blowupat the end of the process as it became an anamorphic releaseprint,” explains Burgess.
He adds that he investigated, and then rejected, using digitaldailies for the live-action sequences. “I felt the quality wasnot good enough to judge what I needed in my dailies to justify theexpense of getting the right projector and hardware to set up a roomand make it all work. We tested it and tried a system and synched-upsound, but it never worked to my satisfaction.”
During production, on the other hand, Burgess used several newertools that he says have revolutionized the way action movies areshot.
“Take the [three-axis] Libra III remote head, forexample,” he says. “I’ve used that before, but when we shotSpider-Man, we figured out how to put the Libra head on acomputerized, high-speed cable winch to fly the camera anywhere wewanted it to go. It worked so well I used it again extensively on thisshoot. We also used the Libra stabilized head on the insert car for carchase sequences, which let us do radical crane arm moves with thecamera without putting an operator in a precarious position. The otherthing we have come a long way with is using radio frequency control forthese tools. We operated the heads remotely from a distant location,keeping all our operators safe. That does mean adding a new person tothe crew — a technician to handle radio frequencies. In my case,I use Nick Phillips [of Geo Film Group, Van Nuys, Calif.], who designedthe Libra head.” Phillips won a 1999 Scientific and Engineeringaward from the Academy of Motion Picture Arts and Sciences for theLibra head, which he later refined for Spider-Man.