Monstrous Task

 Exclusive Podcast

Shooting Cloverfield

Filmed on the Paramount studios backlot in Los Angeles, the “Bodega” sequence of Cloverfield opens with a bang as the Statue of Liberty''s head is hurled down the street by an unseen force. Clouds of dust and debris, both digital and practical, hang in the air and create an apocalyptic atmosphere.
Photos: Courtesy of Paramount Pictures © 2008 Paramount Pictures. All Rights Reserved.

The poster for Paramount Pictures' Cloverfield depicts the Statue of Liberty standing headless, following a monster's attack on New York City. The image echoes the graphic of the severed head of Miss Liberty that adorned the movie poster for 1981's Escape From New York — an image that made a lasting impression on Cloverfield producer J. J. Abrams. But Abrams vowed that in his movie — which follows a group of friends trying to escape Manhattan's destruction — Lady Liberty's head would roll.

This signature moment was one of 160 shots created by the London-based effects house Double Negative (DNeg). “Any time you see destruction, that's our work,” says CG Supervisor David Vickery.

“We had to build a very, very high-resolution 3D model because it's visible in full frame for several seconds during the movie,” Vickery says of the sequence where the statue's head physically rolls down the street.

DNeg's 2D Sequence Supervisor Ben Taylor did such a convincing job that Cloverfield execs later asked to borrow the actual statue model for the film's premiere. “It was a nice compliment,” Vickery says, “because there never was a model on set. It was only done in CG.”

While California-based Tippett Studio handled the monster animation(see sidebar), Double Negative's assignment, which Michael Ellis supervised, required building whole sections of Manhattan in 3D CGI — and then destroying them.

A prime challenge for DNeg was integrating the visual effects with the shaky-cam cinematography used in Cloverfield. While Director Matt Reeves and DP Michael Bonvillain shot the live action with professional Thomson Viper HD cameras, the point of view is supposed to be from one character's handheld digicam. “The camera was zooming throughout the film, and we had to work out that zoom in the computer,” Vickery says.

Although the rooftop scene was filmed on a stage at Downey Studios in Los Angeles, it was London-based Double Negative that digitally extended the practical set and added the second tower of the canted building. The addition of gaping holes, falling debris, and sparking power cables served to make the rooftop feel more dangerous.

DNeg used a range of software to track the constant camera movement. “We used PFTrack [by The Pixel Farm], [2d3] boujou, and [Science.D.Visions] 3D Equalizer — though at least half of it was done by hand in [Autodesk] Maya. We had a very good match move lead, Sam Schwier, but this took a few years off his life.

“We'd have five or six takes of continuous actions, which they'd cut up into sections,” Vickery says. “We'd get the first half of take one and little bits of take two, and then another little bit of take one. We'd have to blend it together to create seamless camerawork. Our experience in Children of Men came in very handy in taking different plates and make them feel like they're a single shot.”

The atmospheric cityscapes DNeg previously created for Batman Begins and World Trade Center also prepared the team to handle the simulated explosions, dust clouds, and virtual buildings that Cloverfield required. DNeg employed proprietary fluid-simulation software called Squirt and a volumetric renderer called DNB to create environmental effects, as well as a tool called Windowbox to create scores of Manhattan skyscrapers. “It's a clever method for putting the insides of buildings back together,” Vickery says. “We have simple polygonal geometry, and we just plane our polygonal faces for the windows, but it looks like there's a true three-dimensional room inside those windows.”

Gathering the necessary survey data to model a virtual Manhattan was one of DNeg's biggest tasks. “[The method] was similar to Lidar but more specific,” Vickery says. “A Lidar system would spray a million points in a scene — more than our software could necessarily handle. We'd go out and measure in 3D space very specific sets of points off buildings and street corners and signage. We'd fire maybe 400 to 500 points in a scene, and at the same time, we'd photograph the area using tiled photography. That allowed us to line our photography up to the three-dimensional points that we've gotten, and then we'd rebuild the sets using photogrammetry.”

Fortuitously, Tippett's crew chose a similar approach, and the studios shared survey data. “We'd upload our photos to almost a common server. It felt like we were on the same team — just on different sides of the Atlantic,” Vickery says.

DNeg's biggest action sequences — one in which the monster wipes out the Brooklyn Bridge, and another where the main characters attempt to evacuate Manhattan in Huey helicopters — required complex 360-degree green-screen effects. “The bridge was visible, full frame, for at least 3 minutes,” Vickery says. “The production built a 200ft. section on a stage, but we had to put in the rest.”

And because there was no chance of filming a helicopter evacuation in Manhattan, an intricate composite was necessary. “We had to recreate the entire corner of 40th and Park. The helicopter was shot at an L.A. airfield. As it took off and turned around, we had to add a huge tanker truck flying through the air and smashing into a building,” Vickery says.

Ironically, when the Cloverfield crew was finally allowed to film the actors running through the actual intersection, it was under construction following a steam-pipe explosion. “So we had to paint out that real destruction and then blow it all up again in 3D CG.”

“Things that are supposed to look gritty and real are the most challenging work,” Vickery says. “You have to be prepared to do iteration after iteration, and it's not always easy to tell why something doesn't look real. But there is a right answer that you have to get.”

Digital rocket trails, muzzle flashes, tracer bullets, and other elements were added to augment practical effects for many plates in which the monster rampages through New York City.

The Evolving Nature and Importance to the Story of Cloverfield's mysterious CG monster (so mysterious that, at press time, Paramount was refusing to give out pictures) greatly challenged the visual-effects team at Tippett Studio. The creature was intentionally designed to be only semi-visible for much of the movie, viewed through the lens of a non-professional videographer fleeing for his life. Most of those shots, therefore, had to evoke the notion of one seamless cut for lengthy periods. Plus, the shots involving the creature had to be built on an extremely modest budget and timeline — even as story elements kept changing late in the game.

“Before we could even begin working with the creature, a lot of the work involved assembling multiple plates to create one seamless cut,” says Tippett's Visual Effects Supervisor Eric Leven. “They shot pieces [of shots] with [the intent] of having us [put it all together] using different editorial and visual-effects techniques. So there was this big additional process of putting plates together, and sometimes, those plates didn't match very well. For example, there is a shot where an Army soldier shoots a rocket at the monster. Plate A shows the rocket going off on a certain part of the street. Plate B is a completely different part of the street, and yet we had to make the rocket trail one seamless trajectory. In those instances, we started in editorial to see if we could get away with a soft cut or a three-frame dissolve here or there. If that didn't work, we might try a basic morph, and if that didn't work, we might go to a 3D reconstruction of [that portion of] the city, if necessary. If none of that worked, we would use combinations of all those techniques.”

Tippett inherited the basic creature design from designer Neville Page, who provided 3D illustrations created in Pixologic ZBrush digital painting and texturing software. The company treated those models as 3D scan data, and then rebuilt the monster from scratch in Autodesk Maya. The initial challenge came in deciding how much detail to build into various parts of the monster's body, given the initial plan to keep the monster's details a bit of a mystery.

“We thought we would see him only in medium or wide shots, so we did not need to put in complete facial rigs,” Leven says. “But, of course, as the movie went on, filmmakers kept adding more shots, and at the end of the movie, we ended up having a huge shot that is completely full frame, full CG. So we had to keep going back and adding things as we went along. Therefore, toward the end of the project, we ended up doing a muscle simulation and adding a facial rig and subsurface scattering.”

According to Leven, the creature was designed to have a translucent, “almost albino quality” — a white monster, probably from the water, with very little exposure to sunlight. That presented Tippett with another problem: the fact that virtually the entire movie takes place at night in a city that has little electricity or light.

Thus, much of Tippett's labor involved tweaking the creature's color and skin texture and giving him a sense of scale. In particular, the addition of parasitic creatures that cling to the monster helped address the scale issue. The team also brought the camera closer to particular places on the creature's skin to let audiences see veins and pulsating blood clots just under the translucent surface.

“The parasites clinging to him — we treated them like barnacles,” Leven says. “When you eventually get to see how big those parasites really are, when they fall off him, it tells you how big he is in comparison to them — the creature is 300ft. tall, so humans can't interact much with him. So filmmakers had the parasites cling to him and then drop off — they are about the size of Rottweilers.”

Leven says that sequences such as the one in which the monster stampedes down a New York street while a B-52 drops bombs on him and the final, full-frame shot of the creature — supposedly shot by a plummeting video camera after the operator is eaten by the beast — were incredibly complicated to animate, match-move, and composite because of the wild camera approach to shooting the movie.

“The camera moves are insane, and so the match-move artists were the unsung heroes of the whole thing — led by [Match-move Supervisor] Devin Breese,” Leven says. “It wasn't that long ago that we were telling filmmakers that at least 50 percent of their shots had to be pan-and-tilt. And here, just a few years later, it's a complete free-for-all — total freedom for filmmakers. We used inhouse tools for that work, but it was largely done by brute force of hand. There are lots of good [off-the-shelf] software tools around now, but I think most people would agree there is nothing to match the eyeball of a good match-move artist.”

Tippett artists used Maya for animation and lighting, proprietary software for match-move work, Apple Shake for compositing, and Pixar RenderMan for rendering the creature. Chris Morley led the compositing team, Steve Reding was lighting supervisor for the creature shots, Tom Gibbons was the beast's animation supervisor, and Peter Konig was Tippett's art director on the project.
— Michael Goldman