Do you call a film with more than 1000 visual effects a visual-effects film? Probably not, if you don’t notice the effects. That's the case with director Neil Burger’s Divergent. “It’s a big visual-effects film, but it’s not VFX-driven,” says Greg Baxter, visual effects producer. “The effects are seamless and invisible."

 

But they were not easy.

 

With Method and Scanline the lead vendors, visual effects artists supported set designers, actors and stunt doubles with digital set extensions, digital environments, CG props and vehicles, and digital doubles as needed including a zipline sequence, establishing shots, a train sequence, and a complex group of shots that put the lead character inside a mirrored room. 

 

The film's protagonist is a young girl named Beatrice, and later Tris, played by Shailene Woodley. She has reached an age where she must choose to align herself with one of five factions in her dystopian future Chicago. To do so, she’s zapped with a substance that sends her into a simulation designed to show her the faction that fits her best. Born into the Abnegation faction, she has shunned vanity, so her simulation stands her — in her mind’s eye and the audience’s view — in front of a mirror. 

 

She looks to the left. That wall and another behind her become mirrors sending an array of reflections deep into infinity. She turns again and the wall behind her becomes a mirror.  All the walls are now mirrors and she faces a kaleidoscope of reflections. 

 

She walks around, breaking the camera plane, and the camera pans around her – the camera doing a 90 as she does a 180. She looks around at the reflections of herself and then comes back. Then the camera is on a two-shot of Tris and herself with reflections into infinity. One reflection, the one closest and largest reaches out her hand, touches Tris on the shoulder, and tells her to choose. Then suddenly an aggressive German Shepherd-like dog appears and threatens her in her thousands of reflections. She calms the dog and it changes into a less furry, more puppy-like canine. She backs away from the dog and another image of her backs away from another dog. They pass each other and lock eyes. They’re facing each other’s dog. Then the dog attacks the image of Tris running away. Tris jumps on the dog to stop it – we see this happen in all the reflections. And, the simulation ends.

 

So how do you film such a thing? 

 

“We did it with no mirrors,” says Jim Berney, visual effects supervisor for Divergent. “We did it all in camera, so to speak. It was the most interesting, hardest, most rewarding and coolest thing I’ve ever done.”

 

 

Director Neil Burger talks about the mirror-room scene in this video from The New York Times.

 

With an Oscar nomination for best visual effects (The Chronicles of Narnia: The Lion, The Witch and the Wardrobe), two BAFTA nominations (Narnia and Harry Potter and the Sorcerer’s Stone), and three VES nominations (Eagle Eye, I Am Legend, and Narnia) to his credit, Berney knows whereof he speaks.

 

“I had talked with Neil [Burger] briefly on the phone about the city,” Berney says. “And then at my first meeting with him, he tried to explain the mirror room to me. I went through a bit of ‘What do you mean?’ and then it clicked. I said, ‘I get it. I can do that.’ But when I was driving away I realized I didn’t know how.”

 

Berney had production designer Andy Nicholson build a small mirror box and he put a doll inside. “I stared at that box for an hour a day for a month,” Berney says. “I acted out the scene in a conference room over and over.”

 

Then he worked with previs artists from The Third Floor to map out the sequence. “People would say, ‘Just get two mirrors and you can do this.’ But I would show them that all the images were contaminated with cameras and camera people. I think when you watch it, it will seem like they shot it. But there’s no way to shoot it.”

 

 

The previs artists created a plan using four to six cameras, each positioned to capture a reflection angle. “We needed all those camera reflection angles. You see her from the front, the side, the back, the other side. All the reflections had to be the same take. And, eventually the room becomes an octagon.”

 

Once Berney had a basic previs that told the director’s story, he sent it to Method. “We got the scenes from The Third Floor, started looking at them, and went, oh wait a minute,” says Matt Dessero, visual effects supervisor at Method. “The room just changed shape. Tris would move in the first reflection plane closer and farther away. And, even the bends of the reflections in the distance were art directed.”

 

Working from the previs, artists at Method created a techvis in Maya, a schematic that showed the position of each of the cameras in each shot. “There were 35 shots in the mirror room, and most had six cameras,” recalls Dessero. “Twenty-six were multiple reflection shots and nine were "infinity room" shots. We plotted roughly 136 camera setups. We had a schematic for every shot.”

 

Using custom code, the team exported the data from the Maya techvis scenes, calibrated a robotic Leica Total Station Survey Head, loaded the data into the survey head, and laser plotted the 136 camera positions onto a 100 x 80-foot green-screen stage. 

 

“Basically, we extrapolated the cameras from the 3D world and put them back into the real world,” Dessero says. “We built a grid on the floor to know the origin for each shot, and we had the schematics that told us where everything was. “

 

 

Two days before the shoot, the Method crew applied tracking markers to the 50-foot-tall green-screen curtains and to the floor. The day before, they tried a test shot. “We had Jim Berney and the full camera and lighting crew, Alwin Küchler the DP, Len Levine the gaffer, and Jim Shelton the key grip,” Dessero says. “Lynn and Alwin had built a complex gridded structure so they could light Tris in flat lighting. We couldn’t have light on the stage. Everything had to be done from above.” 

 

On set were the Alexa cameras, up to six per shot, plus one or two Canon 5D Mark III witness cameras. Each camera had its own monitor with a live video playback outside the green screen. Each camera operator had a still image printed from the techvis. “That showed them what to frame,” Dessero says. “In addition, each camera operator had a map that showed what the shot should look like. We also marked the start and end paths for the actor and even the dog. We couldn’t move the reflection cameras – if you moved one, you had to move all six. So, it was easier to move the actor and the dog.”

 

During the test, and later during filming, the Method crew – digital effects supervisor Blake Sweeney, compositing supervisor Brian Delmonico, CG supervisor Juan Gomez, and Dessero would set up the greenscreen room for each shot. “Brian [Delmonico] would place a cone for every camera,” Dessero says. “Then, he’d set these little placards that had the camera height, tilt, framing chart, lens, everything.”

 

 “Matt and the people from Method executed it brilliantly,” Berney says. “They created all these precise positions and angles. Then, it was all about the eyeline. That was hard. The camera and reflections worked, but when the eyeline didn’t, it was all off.”

 

Berney first tried using his fist to give the actor an eyeline. “That didn’t work for many reasons,” he says. “Somehow in that reflective plane, that angle, you have to put something in space and have it travel. “

 

To accomplish that, Shelton put a dowel affixed with a piece of tape on an “endless” loop. “He pulled one end and pushed the other,” Berney says. “He was basically puppeteering an eyeline and guesstimating. People were rolling their eyes and I told Neil it would take some time. Neil said we’d get it in one take. And we did. When you watch the shot, look at her eyes. She’s looking at herself. It’s magic.”

 

The magic is due also to the post-production crew at Method. The plates arrived with each labeled according to the camera. “We already had the 3D scene from the techvis,” Dessero says. “But we had to track the cameras for the differences.”

 

They started with a layout. “We decided where Tris would be and where all the mirrors should be,” Sweeney says. 

 

Then, they triangulated Tris’s position in 3D space from the plates. “We used the Alexa cameras,” Dessero says. “At the end of every setup we had surveyed every camera so we had full 3D information. We knew where each reflection camera was placed. We placed those and put our CG Tris in the scene. It wasn’t straightforward, but we had all the data we needed. Once we solved the camera, we rendered Tris through the camera and exported data to compositing for every reflection camera.”

 

They decided to base the first five bounces on the Alexa film footage and after that use CG reflections off in the distance. Delmonico placed the reflection cards in 3D space.

 

“In one shot, I placed 108 cards, which gave us five Tris’s deep,” Delmonico says. “From there, we went to 64 bounces. We had thousands of Tris’s in that shot. Early on, we tried using all the Alexa reflections, but it was a lot of work to art direct and composite.”

 

Dessero explains that within the renderer VRay, they could cancel out the first five reflections and take over with CG at the sixth. “We’d render one pass with every single Tris so compositing would have that reference, and then another pass where we removed the first ones,” he ays. “It was a simple ray switch. Early on we thought we could get away with 16 or 32 reflections, but we rendered one shot with 64 and from then on, every shot had 64 depth reflections. Sixty-four ray bounces in depth. It was crazy.”

 

For each of the 35 shots, the compositors had to rotoscope Tris from six plates, and for the early shots, add ceiling, walls, and floor. “Every marker that tracking loved, the painters hated,” Dessero laughs. To help make the sequence seem more organic, compositors added smudges and streaks to some of the mirrored reflections.

 

 “We solved every problem we encountered one at a time,” Sweeney says. “But looking back, it was grisly.”

 

 

For the train sequence, Method artists extended the environment shot in Chicago, adding tracks and creating CG train cars. “If the camera is in tight, the actors are getting into and jumping off a car on set,” Berney says. “But the car didn’t have a top, front, or back. And if you see more than one car, the train is all CG. Method knocked it out of the park. They created the neighborhood based on real-world buildings. And when the actors are running on the track, they roto’d them out of the footage and put them on a high track. It’s phenomenal.”

 

To match the practical half-car on set, the crew used photogrammetry. “We needed to be really accurate,” says Juan Gomez, CG supervisor. “The photogrammetry images also helped with the textures. We had exactly the same details on our CG train as on the practical train.”

 

A rig added vibration when the train moved down the track and animators could dial a rocking parameter between cars up or down depending on the shot. For scenes with miles of train tracks and truss structures, instanced geometry lowered render time. 

 

Method artists also created the establishing shot that opened the film, and built the “Dauntless” compound including a glass atrium. “It was a huge build,” Dessero says. “Some shots were filmed on greenscreen or backlots, some on location. There was a big mix of locations, but at the end of the day, no one will know. Juan and his team put together beautiful invisible effects.”

 

Similarly, artists at Scanline created invisible effects for a zipline sequence. 

 

A stunt girl performed much of the action. “The original idea was to shoot helicopter plates as she traveled on a zipline from the top of the Hancock building to an alley,” Berney says. “In the end, the studio wanted more, so we redid the whole scene with Scanline. We used the plates, but when the camera didn’t match the new action we projected photography onto geometry.”

 

For these sequences and others the studios created, the goal was to always build from photography. “None of the 1000 shots are in your face,” Berney says. “Even the compositing is done so that everything looks like it’s in camera.”

 

That created an interesting dilemma for the director and visual effects supervisor as they thought about how they’d tell their story. Berney has the anecdote: “On my last day on the show I was in the DI suite and Neil said to me, ‘I don’t know what to do with the interviews. I don’t want people to look at this as a visual effects movie.’ And, he’s right. We shot as much as we could. Even the CG is built on photography. So, I said, ‘Well don’t lie. Do what you need to do. And, we’ll represent our work in our world.’”

 

And so he has.