Read in-depth features about how cutting-edge CG was created for various movies making Oscar buzz.
Visual effects studios create CG explosions and virtual enviroments for Angels and Demons.
The clock is ticking on an unstoppable Illuminati time bomb hidden somewhere in Rome. Only one man, Harvard religious expert Robert Langdon, can distinguish angels from demons and find the bomb in time to prevent an attack on the Catholic Church. To do so, Langdon must get inside the Vatican’s archives to discover the path to illumination. But, the Church does not hold the questioning symbologist dear to its heart.
The Catholic Church didn’t hold dear the filmmakers who had also made the movie The Da Vinci Code, either. Based on the controversial novel by Dan Brown, the 2006 film directed by Ron Howard had sparked harsh criticism from the Roman Catholic Church before its release. Thus, well before postproduction had wrapped on that film, Angus Bickerton, visual effects supervisor, began preparing for Columbia Pictures’ Angels & Demons, Brown’s prequel turned into a sequel.
Bickerton says, “As we wrapped up, Todd Hallowell [executive producer, second unit director] said there was every chance we’d be doing a sequel, so because we could have problems, maybe we should photograph the key locations in Rome whilst we were all still together.”
For The Da Vinci Code, the visual effects studio Rainmaker (now CIS-Vancouver) had circumvented the Church’s restrictions and created the interior of Saint-Sulpice by projecting photographs of the interior onto low-res geometry, and then placing actors filmed on a partial set into the virtual environment (see “On Holy Ground,” July 2006). “Todd liked that methodology,” Bickerton says, explaining that the Church had not given permission to film inside Saint-Sulpice. “Using photographs of the actual location worked really well. We had a partial set, but 85 percent of the church was the virtual environment.”
With this in mind, Hallowell and a photographer traveled to Rome, where most of the sequel takes place, and did the “Angels & Demons tour,” as Bickerton puts it. Although none of the post houses used those photos for textures or photogrammetry, the images served as reference material.
“Ultimately, each studio working on the film essentially approached the project in the same way,” Bickerton says. “They re-photographed the environments and derived geometry from online sources and from the photographs themselves.” To do so, teams from the visual effects studios joined the throngs of tourists and shot still images with digital cameras. Although the Church wasn’t as hostile to the sequel as the crew had feared, the filmmakers still couldn’t shoot actors inside the churches, nor could the VFX houses take laser scans of most locations.
“We didn’t pretend to be tourists, we had tripods, but it certainly wasn’t a proper setup for a shoot,” says Graham Jack, CG supervisor at Double Negative (dNeg).
Bickerton selected four visual effects houses to work on the film, the same studios he had worked with on The Da Vinci Code: Double Negative, CIS-Vancouver, the Moving Picture Company (MPC), and the Senate. “After seeing Cloverfield particularly, we decided dNeg was the best option for our big explosion sequence,” Bickerton says. “Once we had chosen them as our primary company, we based ourselves in London.”
Double Negative also created the exteriors of St. Peter’s Basilica and St. Peter’s Square, the crowds that fill the square, a CG helicopter, daytime establishing shots, and exteriors for the funeral shots. CIS-Vancouver built the basilica interior, the fire sequence, and the interior of Santa Maria della Vittoria and Santa Maria del Popolo. MPC worked on shots in Piazza Navona, inside CERN, and the antimatter sequences. And, the Senate handled shots for the Pantheon and inside an underground Vatican vault.
In addition to the four post houses, Bickerton worked during production with compositor Duncan Kinnaird. “We’d send him shots from LA, and he’d comp them overnight with (Adobe) After Effects, using backgrounds we shot or temporary stills,” Bickerton says.
Even though Kinnaird used 2D stills for the composites rather than the 3D backgrounds the effects houses would eventually create, Bickerton found them useful. “It was the first chance to see how the sequence worked and get responses from Ron [Howard],” he explains. And the rough composites were perfect for Howard.
CIS-Vancouver extended the set for Santa Maria della Vittoria here and in the image on the previous page, and composited fire into the image above using photographed elements and CG fire.
“We were quite a low-tech cutting room, cutting at NTSC resolution in the Avid because that’s the way Ron and the editors like it,” Bickerton notes. “They like it low res and rough for cutting because they have somewhere to go, and because they can say it’s a work in progress when they show it to the studio. But what these temps do is give them a picture of what Tom Hanks is standing in front of, which affects the dynamic of the shot and lets them cut for picture content.”
At Double Negative, Ryan Cook led the visual effects team, which created approximately 200 shots and graded another 70. “They had the biggest environment and the biggest workload,” Bickerton says. To build St. Peter’s Square in the Vatican for that sequence, dNeg modelers working in Autodesk’s Maya and Pixologic’s ZBrush created the 284 columns in Bernini’s monumental colonnade topped with 140 statues of saints that encircle the huge, elliptical piazza, two fountains, the 83-foot-tall obelisk, the statues of St. Peter and St. Paul, and the exterior of St. Peter’s Basilica. (CIS-Vancouver built the basilica interiors.) In the film, thousands of people swarm into St. Peter’s Square to wait for the white smoke that announces the election of a new pope.
“We photo-modeled what we could from our photographs using dnPhotofit, our photogrammetric software, and we relied on tourist photographs and aerial surveys,” says Jack. “For textures, we have software originally built for Batman that automatically stitches the bracketed exposures and tiles that we shot with long lenses into panoramas. When we didn’t have enough detail for the baroque architecture and statues, we relied on our texture painters and sculptors.”
The bigger problem was lighting the square for day and night. At night, lights point up to the top of the colonnade, floodlights point down, and numerous lights illuminate the basilica façade. “It was the most complex lighting we’ve done,” says Jack.
DNeg uses Pixar’s RenderMan, and to help optimize the shadow generation for the hundreds of light sources, the team rendered multiple passes using RenderMan’s point-cloud occlusion. “We rendered the floor, the colonnade, and the basilica into different point clouds,” Cook explains. “And then to calculate the ambient occlusion and color bleeding, we combined all the point clouds into one using a little command-line tool that read in the multiple passes after we baked them. So, if the lights had to change in one area, we needed to re-calculate only that layer.”
Double Negative extended the set built for St. Peter’s Square (left) using 3D models and projected textures. Artists then filled the set with crowds they created and animated using two particle-based techniques for the shots in the final sequences (right).
To create the crowds in the square, the team applied two techniques. Particle-based sprites created with Maya and RenderMan put extras, shot on a greenscreen stage, into the background. Artists placed these sprite-based crowds by painting areas on the ground plane. When people needed to move in particular ways—especially when the shock wave from the explosions hits the crowd—the crew used a solution built around Side Effects Houdini’s particle system and its channel operators (CHOPs).
“Every particle has a metaball associated with it,” Cook says. “They move based mainly on how crowded the area is around a particle; the system can find the nearest particle and plot a course to avoid other particles. Because the system knows the position of all the particles, it works out, based on velocity, how to blend motions and trigger actions. It matches the size of a step to the distance traveled, and blends between walk speeds.” Motion capture provided the animation clips, and a RenderMan skinning tool provided bodies for people.
The explosion is a combination of practical effects and a fluid-dynamics simulation created at dNeg. “We wanted a believable explosion in a style that echoes things people have seen, but something that no one has ever seen before,” Bickerton says.
At first, the dNeg team created nuclear explosions using Squirt, the studio’s in-house fluid-dynamics solver, but Howard felt that was too cataclysmic. “We changed gears to develop something more astronomical and bathed in light,” Jack says. The resulting explosion starts with pyro footage, then becomes a fluid-dynamics simulation that contracts to a point and expands back out.
“To do the final effect, we used a combination of fluids rendered in our volume renderer, procedural effects, and matte-painting elements, all composited together in (Apple’s) Shake,” Cook says. “Our fluid solver allows you to compress or expand, so we could suck the explosion back to a point.”
Path of Illumination
Except at the very beginning, the entire movie takes place one night in Rome, with Langdon trying to deduce where the Illuminati would execute four kidnapped cardinals, one after another. Each location represented a primordial element, and CIS-Vancouver helped create the interiors of two: The Chigi Chapel in the church of Santa Maria del Popolo, where Langdon arrived too late to save a cardinal branded with an Earth symbol from death, and the church of Santa Maria della Vittoria, where he arrived too late to save another cardinal from immolation. Actors were filmed inside detailed sets for both these church scenes. CIS extended the sets and created the fire that burned the second cardinal, using a mix of practical and CG elements. The majority of CIS’s work, though, centered on the enormous St. Peter’s Basilica.
CIS used NewTek’s LightWave to extend the Santa Maria del Popolo set, but built everything else, including St. Peter’s, with Maya. They rendered the models with Mental Images’ Mental Ray, and composited the shots with Shake. Visual effects supervisor Mark Breakspear led that studio’s work, as he had for The Da Vinci Code.
The cataclysmic and climactic explosion starts with pyro footage, changes to a fluid-dynamics simulation that contracts to a point and then, still using fluid dynamics, expands back out.
To prepare for the studio’s work, Breakspear and his team took nearly 10,000 photos in Rome using Canon EOS 22-megapixel cameras equipped with 8g Flash cards. “We used those to build our CG models,” he says. “They became the basis for previs, and we improved and improved them until we could use them in the movie.”
The previs helped Howard with pacing, and DP Salvatore Totino with framing for shots in the huge expanse. On set the filmmakers had surrounded a replica of the base of Bernini’s baldacchino, the 90-foot-tall bronze canopy located beneath St. Peter’s dome, and a section of the floor with greenscreen. CIS built the rest of the baldacchino, the 138-foot-diameter dome (which rises 452 feet above the street), and the entire 694-foot-long interior in CG. The building, designed in large part by Michelangelo, covers 5.7 acres.
“The previs renders were primitive, of course,” says Karen Ansel, CG supervisor. “But, our model duplicated the entire length, so it helped Sal [Totino] know what chapel the actors might be going past.”
Once the crew got rough cuts back from shots for the basilica sequences, they knew what they’d need to render and what textures they’d need. “We went back to Rome knowing the camera angles,” Breakspear says. “We could stand almost in the same spot and take all the pictures we needed, so it was a very efficient process. Our technique is that you work out where the camera is and take pictures from there, from the end position and all the way along the in-between bits so the texture can adjust for different camera movements.”
CIS-Vancouver rendered the interior of St.Peter’s Basilica using as many as 20 layers to create and more easily change any mood without weighty re-rendering, whether in daylight (Left) or at night. At right, the floor is a 3D model with detailed shaders and texture maps.
Back at CIS, artists readied the photographs for projection onto low-res geometry by painting out the tourists and bringing the images back to neutral lighting. “When you take HDRIs, if you have a shot with strong sunlight hitting a column, you can bring back the detail using one of the other exposures,” Breakspear says. “It’s a painstaking process. But, the meat of our work is in refining those textures to be beautiful and crisp.”
The lighters used three setups in Maya and Mental Ray to render for daytime sequences, dark and somber night scenes, and nighttime action sequences. “In addition to the regular ambient occlusion, shadow passes, depth maps, and specularity layers, we rendered passes with 3D lighting that worked with the projection maps to create lighting effects,” Ansel says. That meant the lighters could create areas of dark and light, cool and warm, and the falloff without weighty re-rendering. And, if the look changed, the crew could re-render any of the 16 to 20 layers sent to compositing.
In addition to the projected textures, CIS artists modeled the basilica’s marble floor in 3D using detailed shaders and texture maps to create the aged tile inlays and to reflect the baldacchino, itself a fully detailed and shaded 3D model. And, light beams created with volumetric lighting and dust motes added to the atmosphere. “Technically and aesthetically, we wanted to support the suspense in the story,” Ansel says. “And, we wanted to keep the basilica as inspiring in the film as it is in real life.”
Similarly, MPC used projected textures to create shots of the actors in the Pasetto, a secret passageway disguised as a Roman aqueduct. “We were allowed on the Pasetto, so we photographed it using a motion-control tessellator head and four digital still cameras on tripods,” Bickerton says. “MPC built limited geometry to project that imagery onto. And, the Senate used photo-tiling for the Pantheon, which has some fully CG shots.”
On set was a replica of the base of Bernini’s bronze baldacchino that rises 90 feet from the floor, and a section of the floor. CIS-Vancouver built the rest of the baldacchino and the entire 694-foot-long interior of St. Peter’s Basilica using CG models and projected textures.
“Hopefully,” Bickerton says, “the effects in this film are as invisible as possible. My aim was to be photographically real.”
And it is. After all, even though they had to film many of the scenes on soundstages in LA, they did shoot the film in Rome in a way, albeit with still photos turned into photoreal locations, thanks to the science of computer graphics and the skills of CG artists.
Barbara Robertson is an award-winning writer and a contributing editor for Computer Graphics World. She can be reached at BarbaraRR@comcast.net.