The following is an extract from my article on Spectre, published in Cinefex 145, February 2016

In Spectre, British Secret Service agent James Bond (Daniel Craig) pursues a trail of clues that eventually leads him to the heart of a global terrorist organization headed by Franz Oberhauser (Christoph Waltz). Ably assisted by colleagues ‘M’ (Ralph Fiennes), ‘Q’ (Ben Whishaw), and Eve Moneypenny (Naomie Harris), as well as new companion Dr. Madeleine Swann (Léa Seydoux), Bond crosses three continents on his way to an explosive finale in which he confronts not only the dastardly Oberhauser, but also ‘C’ (Andrew Scott), treacherous chief of the new Joint Intelligence Service.
Emulating Bond’s reliance on trusted allies for support, producers Barbara Broccoli and Michael Wilson gathered round them a closely-integrated team of familiar faces to help bring the 24th ‘007’ film to the screen. Having cut his teeth on Skyfall, director Sam Mendes returned to helm his second Bond movie, supported by special effects supervisor Chris Corbould and visual effects supervisor Steve Begg. Dennis Gassner reprised his role as production designer, while newcomer Hoyte van Hoytema took up the reins as director of photography.
Eschewing the crisp, digital aesthetic employed by his predecessor Roger Deakins, van Hoytema opted to shoot Spectre on 35mm film using anamorphic lenses. “Hoyte was after a very cinematic, romantic look,” said Steve Begg. “That had to be reflected in the visual effects. We had a few tracking and resolution issues — anamorphic is inherently just a little bit softer on the edges, particularly on film — but nothing really profound.” Begg shadowed the main unit throughout an extensive shoot at locations in Mexico, Italy, Morocco, Austria and England, and on soundstages at Pinewood Studios in the U.K. “My main task was to stick with Sam and Hoyte,” Begg commented. “I think it paid off, particularly in post. We ended up talking in shorthand because I had been there when the shots were being made, and I knew exactly what they wanted.”
Under the guidance of Begg and visual effects producer Leslie Lerman, supported by associate visual effects supervisor Andrew Whitehurst, Spectre’s 1,540 visual effects shots were divided more or less equally between Cinesite, Double Negative, Industrial Light & Magic, MPC and Peerless Camera Company, with a small amount of additional work overseen by Stuart Bullen at BlueBolt. “The film broke nicely into segments, so we chose the work appropriate for each company,” said Begg. “There was very little overlap between them.”
According to tradition, the main titles of a Bond movie are mini-productions in their own right, filled with surreal moving images set to suitably moody music. The Spectre title sequence, the seventh with director Daniel Kleinman at the helm, is no exception. Kleinman worked with long-time collaborators Framestore, reuniting the team behind previous 007 titles including Casino Royale and Skyfall. “Danny came up with the initial ideas,” said Framestore creative director William Bartlett. “Then we talked about the themes in the film, and the kinds of metaphors we could use. For example, we knew there was going to be an octopus, and a surveillance theme, so we suggested, ‘What about having an eye, only the iris is made of tentacles?’”
The octopus — symbol of the terrorist organization Spectre — became the sequence’s dominant image. Early concept development focused on the creature’s appearance. “The more you look at octopuses,” said Bartlett, “the more you are blown away by just how weird they are. Ours is partly based on a living creature, only stylized to get the look we wanted.” Skin color and texture were primary concerns. “We started with something very natural-looking,” Bartlett remarked, “but it wasn’t sinister enough. So we went with a dark, slimy look more in keeping with the criminal organization vibe.”
Preliminary lighting and animation tests presented the octopus in an underwater environment. A more menacing creature emerged as the watery cues were sidelined. Throughout, animating the tentacles remained a challenge. “I have never worked on a job where the rigging was going on so late,” remarked Bartlett. “An octopus tentacle does not exactly move from the tip, nor does it exactly move from the base. It stretches at certain points and not at others. It was very tricky to rig.”
An early version of the sequence featured dancers sporting octopus skin textures. “It looked very beautiful,” Bartlett commented, “Octopuses have this incredible texture, but when you see it on a human being I am not sure you would recognize what it is.” Ultimately, cephalopod skin gave way to flames, with practical fire effects composited into a bluescreen shoot of Daniel Craig and his Bond girl entourage. Particle effects added drifting embers and individual flames trailing from the dancers’ hands.
Octopus and dancing girls finally unite as Bond plunges into an ink-filled abyss. “We shot the ink in our cloud tank,” Bartlett revealed, “We squirted in various types of paint, and used those elements to make the environment. We generated specific bits of the girls’ hair using procedural fluid dynamics in Houdini.”
Attempting to evade the ubiquitous octopus, Bond experiences visions of former friends and enemies. The ‘broken glass’ look of these shattered memories repurposed live-action from previous films. “Chaos Group gave us beta test licenses for their new V-Ray plugin for Nuke,” Bartlett said. “That enabled us to do proper refractions and reflections within a 2D system, which handled the live-action more efficiently than if we had done it in 3D.” Additional treatment aided the blending of footage shot in multiple formats. “Some of the live-action was on film,” Bartlett noted, “and some was on digital. Some was shot underwater. It was all lit differently, so we found a stylized look to bind it all together.”
As the sequence concludes, the iconic Spectre ring fills the screen. The Framestore team were keen not to draw parallels with another memorable piece of movie jewelry. “We had a version of the shot where we traveled through the middle of the ring,” recalled Bartlett. “But it seemed a bit too Lord of the Rings. So we ditched it.”
Spectre opens with an attention-grabbing ‘oner’ in which the camera follows James Bond and his companion Estrella (Stephanie Sigman) through crowds of revelers celebrating Mexico City’s Day of the Dead festival. In an apparently unbroken take, they enter a hotel and take an elevator to a room on an upper floor. Bond exits the window, crosses the roof and targets a sniper rifle on the participants of a secret meeting being held in an adjacent building. Lasting some four and a half minutes, the bravura shot was photographed with spherical lenses to afford more flexibility in post, and combines multiple live-action plates stitched seamlessly together by a team at ILM’s London office.
IO Entertainment spent six months developing previs for the opening shot, with available camera movements dictated by an early decision to shoot the opening and closing segments using 50-foot SuperTechno cranes. “We managed the creative choreography and composition of the shot,” said previs supervisor Brad Blackbourn of IO Entertainment. “We also determined logistics, right down to how long the camera track needed to be, whether we needed to build scaffolding for the grips, and which lamp posts would have to be removed on location so the crane arm could swing through its arc.”
The shot begins as Bond and Estrella make their way down a busy street. ILM made numerous fixes to the original location plate, tracking carnival masks to the faces of extras caught looking at the camera and adding digital crowd extensions. As the couple turn into an arched entrance, the camera lingers briefly on a poster to disguise a soft transition to the second plate, which was shot inside a hotel elsewhere in the city. This and other transitions were implemented by reprojecting the live-action onto geometry, which was then warped to get as close a match as possible between the end of one plate and the start of the next. The performances of the actors dictated the exact timing and position of the blends, with the majority of the warping being applied to the scenery around them.
The second transition occurs as Bond and Estrella enter the hotel room, which was built as a set at Pinewood, with bluescreens placed behind the windows to allow the ILM team to composite in exterior plates of Mexico City. A re-time of the footage created the illusion of a whip-pan as Bond moves to the window; a third transition occurs as he exits through it. Footage of Daniel Craig walking along the roof was captured in Mexico as a single plate, with ILM performing rig removal, adding a gap over which Bond jumps, and compiling replacement backgrounds from multiple sources.
Reaching the end of the roof, Bond takes up a sniping position with his rifle aimed at the window of an ornate building dubbed the ‘Fru Fru,’ which ILM developed as a digital matte painting based on the building that was shot in Mexico. “The Fru Fru was a bit of a hodge-podge of different bits and pieces,” said ILM London visual effects supervisor Mark Bakowski. “We used lidar and total station data from the location for the base model, but we reshaped it and topped it up with a couple more floors. The room behind the window was a studio set filmed separately at Pinewood.”
In its final form, the opening shot ends as Bond draws a bead on his target. However, the original intention was to continue the one-take conceit throughout the subsequent action, during which an unexpected explosion causes the Fru Fru to collapse. At one point it was proposed that this — along with the later demolition of the MI6 headquarters — be achieved using large-scale miniatures. “I come from a miniatures background, so I always look for ways to use them,” remarked Steve Begg. “They were very successful on Skyfall, and Sam kept pushing for their use. But, when we found out about this elaborate camera move, I immediately thought, ‘Forget that — it’s going to be CG.’” The crew conducted tests, but ultimately production used no miniatures in the film.
Bond’s gunfire triggers an explosion which causes the Fru Fru’s façade to fall forward and demolish the rooftop where he is crouched. The complex stunt was shot full-scale on the Pinewood backlot, with the chain reaction event recorded by 46 cameras. Previs was used to determine not only camera angles, but also the basic geometry of the physical rigs that would be required. “Once the previs was locked down,” recalled Brad Blackbourn, “we looked at it with Chris Corbould to determine what pieces would actually require building, and to define the approximate pivot points of breakaway sections of the roof.”
A special effects team led by Chris Corbould first constructed the front of the building in its final, exploded state, then reinstated the façade with intact masonry made from soft foam, ready to be blown out by pyrotechnic charges. Special effects then built the roof onto which the resulting rubble falls in three breakaway pieces. “We had a section approximately 30 feet by 12 feet that came down a track and impacted with the roof,” explained Corbould. “That took out the end third. The next third was triggered to release once that first breakaway piece had gone. The last third, which Bond runs up, hinged down on hydraulic rams. It was all very closely timed, one event after another — and it took months and months of planning and testing.” Although the entire rig was capable of being reset, the complete event was captured successfully on the first take. Multiple takes on just the final hinged section allowed the stunt team to perfect Bond’s scramble over the disintegrating roof.
Using the resulting live-action as a template, the explosion and collapse were re-created in digital form. “Chris’s rigs helped us work up a believable camera move with the stuntman running over the collapsing roof,” said Steve Begg. “Once Sam and the editors were happy, we gave those elements to ILM, who pretty much replaced everything.”
ILM’s digital building was constructed in Autodesk Maya and textured using The Foundry MARI and Adobe Photoshop. “We built it up using individual components, then applied material-based destruction to them,” Mark Bakowski explained. “At the base of everything was keyframe animation, with the building pre-modeled to bend and buckle appropriately. Once we had the basic collapse blocked in, we dressed in multiple secondary simulations for dust, rebar, rubble, glass, and so on. We could re-block it all very quickly — it was very light on its feet.” Effects were integrated in Side Effects Houdini, with final renders output through Pixar’s RenderMan.
ILM also married a digi-double head of Daniel Craig to his stunt double’s torso, which ended up being the only part of the live-action retained from the plate. “It was an interesting challenge keeping the CG head looking like Daniel Craig,” Bakowski observed. “The minute we put too much of a performance into him, he started to look particularly un-Daniel-Craig-like! We slightly over-emphasized certain features like his chiseled cheekbones. We would keylight those a little bit, slightly contrary to shot lighting, to give him that classic Daniel Craig look.”
Text copyright © Cinefex 2016. Reprinted here with permission.