Falling Ice, Enjoyable Destruction, and Ninja Fights at 30 MPH

The boys based on Hasbro's G. I. Joe toys are back in town, fighting two enemies now — Cobra and threats from within the government. Channing Tatum returns as Duke, the head of the elite unit in G. I. Joe: Retaliation. Dwayne Johnson is Roadblock, Duke’s second in command. Bruce Willis makes a cameo appearance as General Joe Colton, for whom the Joes are named. For this film, the Joes stuffed the last film's power suits in the closet in favor of a first-person shooter and hand-to-hand combat style, battling Cobra and the Red Ninjas in set pieces around the globe. Director Jon M. Chu of Step Up fame choreographed the high-caliber action in the Paramount Pictures release. James Madigan was the production visual effects supervisor.

We talked with Bill George, who supervised Industrial Light & Magic’s part in the G. I. Joe action. ILM artists created 220 shots for two over-the-top sequences: a fight on a cliff face in the Himalayas and subsequent avalanche, and the destruction of London. George is an ILM veteran who started his career as a model maker (Star Trek II: The Wrath of Khan). He received an Oscar in 1987 for his work as model shop supervisor on Innerspace and an Oscar nomination in 2004 as a visual effects supervisor at ILM for Harry Potter and the Prisoner of Azkaban. When not supervising visual effects at ILM, George creates content for his website, scifiairshow.com, whose stated purpose is “to preserve and promote the rich and varied history of sci-fi/fantasy vehicles.”

StudioDaily: Let’s start with the Himalaya sequence.

Bill George: They shot the majority of the film in an old NASA facility outside New Orleans originally constructed to build the first stage of the Apollo mission. I came in to ILM on a Tuesday and read the script. They awarded the show to us on Friday. On Sunday, I was on a plane to New Orleans. Our job was to create the backgrounds, which were 100 percent digital. And to sweeten things.

They had built a giant angled set [with a] flat green screen made of plywood so they could swing the actors on ropes to do the ninja battles. It was good guys vs. bad buys fighting over a body bag. It felt like spiders fighting over a fly in a cocoon. We tried to get as much as we could on set, but the stunt riggers were swinging at 30 mph, so it was scary. We made digital doubles and sometimes replaced an arm or a leg [or] sometimes the whole character based on the movement John Chu wanted.

John Chu had done Step Up and Step Up 2, so I wondered why he’d go from that to G. I. Joe. But after one day on the set dealing with ninjas, it totally made sense. The ninja battle was all about choreography and motion. It has no dialog from beginning to end. Just like dance.

How did you create the backgrounds?

They were all digital matte paintings. James Madigan shot a lot of reference material in British Columbia from a helicopter, taking textures as he flew along. We triangulated the cliff face from those texture photographs and did a high-resolution model. Then we used the same photographs to project the textures and created a virtual location. The majority of the action had to be at the highest resolution because it was close to camera.

That sounds straightforward. Were the backgrounds fairly easy to do?

No, it’s never easy. There were a lot of technical and creative issues in making it look real. There was a production decision to shoot it as if they were in the shadow side of the hill. That’s a great thing if you want to flop the shot and go left to right. But the lighting was flat on the characters. We had to play them as though they were in shadow, but we tried to get the sunlight close to cheat it, to make it look like they were outside, not on a set. Outside, the contrast range is different. There’s more depth of field. You can’t get detail in shadows and the bright clouds are usually blown out. Those are the imperfections that make it look real.

Were there other technical and creative issues for the backgrounds?

It was a granite, grey cliff face and there wasn’t much of interest on it. No plants. Very little snow. Our digimatte supervisor, Johan Thorngren, and I looked at reference from Yosemite and saw that even though a cliff was granite, it has reflectivity. It has a sheen. So we tried to incorporate that into our rocks. You’ll see reflections play across it. It’s subtle, but adds a lot of life. We did that with a separate render pass.

Watch the video to see behind-the-scenes footage of the green-screen shoot, followed by finished clips and a longer excerpt from the cliff fight.

Is the avalanche a total rigid-body and particle simulation?

We modeled a few big pieces that we hand-animated and then simulated all the other pieces so all the collisions were correct and had gravity. It’s like any creative endeavor. You draw, block in the big shapes, and then refine.

It was almost like a calving glacier. I think falling chunks of ice are scarier than falling snow, which is fluffy and soft. So you see these giant chunks of ice fall and hit the ground, break into finer pieces, and atomize. We had a variety of sizes that added up to the debris. Live-action elements. Lots of particles. Mist. If you just have particles, it looks granular, grainy, so we used Plume [ILM proprietary simulation software]. It creates something very fine, like smoke. My philosophy is that if you mix a variety of techniques, you fool people more. Our eyes and brains pick up patterns, so I try to keep it as messy as possible.

Speaking of messy, describe the London destruction concept. 

In the shot, Cobra launches a weapon from a satellite, like a meteor the size of a trash can, that hits the earth. The studio wanted something unique, not a nuclear explosion. They wanted something more like shattering earth, like plate tectonics. So we created London and shattered it like a piece of glass. You see Parliament, the London Eye, the customs house where you buy your ticket for the London Eye, the Thames river. As the plates tip and turn, you see water going over the edge. The part in the trailer is like an ILM hit list. You get to enjoy the destruction a lot more in the film.

It sounds like a lot of simulation.

Every department was involved. Modelers, painters, digimatte artists, animators, compositors, lighting TDs, rendering. We had particle simulations, rigid-body simulations, dgital matte work. We pulled out all the stops. The great thing was that we had a team of people coming off The Avengers and Battleship, so they were at peak efficiency and totally up on destruction. It was like the Justice League. They were all there. They all had their super-powers. It was awesome. We had to do the work quickly, but everyone was on top of it. We were really lucky.

Did you work from plates?

They sent us helicopter plates, shots flying over London, and said, “These are your background plates.” We didn’t use the plates except for the choreography of the camera and lighting on our backgrounds. It was easier to completely create and destroy London than to blend in. We did a match-move of the plates they liked and then it was easy to figure out the ground plane and where the buildings were. I’d ask John, "Where do you want it to hit?" and then forge ahead.

It was the same as with the avalanche. If you rely on the computer to do the simulation, it does what it does. So I like to take the big pieces and hand animate them. Do the big blocking and let the computer do the small pieces. We got the timing down and then put in the details. I find that’s a more direct way to get to your goal.

This was the first time I’d used Nuke’s deep compositing. We needed to give it a try because we had such a terribly complex scene with so many bits and pieces. It really helped us out.

There’s a lot of simulation work in this film. Was it difficult to get approvals in process?

The software is getting more controllable and John Chu was a dream. A lot of directors are intimidated by CG, but he is a young guy. He grew up in the age of CG and had no fear at all. I’d say, “I’m thinking we should do this in CG,” and he’d say, “Great. I love that idea.” A few years ago, you’d say, “CG,” and the director would say, “It’s not going to look fake, is it?”

I try to make these film fun times for the director. I want it to be like Christmas. I want them to be rubbing their hands, saying “Oh boy, what have you got for me?” What we have is unique and special and I want them to see that.

Well, the London destruction is cool. And it goes by quickly.

You’ve just described my life.

What did you learn from working on this film?

Sometimes that’s hard to quantify. Visual effects supervisors are lucky because we can interface with lots of departments. We can see how the second-unit director and stuntpeople approach shots, how the editors solve problems. I think if I had to boil it down, though, I think the thing we spent the most time on was the animation, the movement, getting the weight right. Paul Kavanagh [an Oscar nominee for Star Trek] was our animation director at ILM. We would present our work to John [Chu] usually over CineSync. He had a camera on him and would act things out. It was a challenge — people flying through the air swinging swords at each other — but John was very clear about his direction. A motion that is two frames different can make all the difference in the world.