VFXWorld goes undercover with Marc Forster and Kevin Tod Haug to explore the "editorially-driven visual effects" on Quantum of Solace.
Indie director Marc Forster (The Kite Runner), a stranger to both franchises and action/adventure, approached Quantum of Solace as a direct sequel to Casino Royale that delves more deeply into the psyche of James Bond (Daniel Craig), following the love, loss and betrayal that he experienced in the origin story. Indeed, the very title, culled from an off-beat Ian Fleming short story, not only alludes to the secret organization of baddies, but also the sense of closure that Bond craves.
Not surprisingly, the Swiss-German applied his penchant for crafty precision and visual style, inspired by the modernistic production design of Ken Adam. Quantum is certainly a leaner and meaner Bond for the 21st century. The director naturally tapped several of his long-time crew members, including Visual Effects Designer Kevin Tod Haug. Working with Double Negative, Framestore, Machine FX, MPC and MK12, Haug oversaw 900 vfx shots within the long-standing Bond tradition of depicting realistic-looking action.
"I pretty much did what I've always done for Marc," Haug suggests. "There is a certain amount of polishing that Marc needs out of the visual effects department because he's developed a recognition that a certain amount of what is too expensive to do on the day could be managed later: 'I know I have the raw material here and ultimately I want to see how it cuts before I deal with some of these issues.' And unlike a lot of directors who go back and do reshoots, he just fixes them. So the stuff tends to be 98% there. There's just that little bit of tweaking to make things look better. Since Stay, we usually have our own in-house level of compositing so that we can do what I call 'editorially-driven visual effects': splitting performances and retiming things, taking a performance from one scene and putting it in another in order to shorten a scene or transplant a scene or a performance.
"But the thing about working on a Bond movie is that neither of us had ever done a giant action movie before. And so it's the pure scope of doing action and the number of things that you have to pull off. We had a conversation early on where I told Marc that I thought he was intentionally crazy to do it because they only had 12 weeks of post, and we'd never done anything that quick -- he'd never done a director's cut that fast. But I think we all agreed that the opportunity to be the first [predominantly American crew] ever to do a Bond was too hard to turn down. I am the first vfx designer (American supervisor for that matter). The rule was don't do anything that you're not 100% certain won't look good, don't get experimental, don't over reach, just do what needs to be done and do it as well as possible. It was a sort of rear-guard action from day one to make sure that we didn't end up with 12 weeks to go and some horrible mess to sort out."
Forster, who dealt with the four elements (earth, water, air and fire) in conceptualizing the action, elaborates: "We only had six weeks to cut the movie and then another few weeks for the sound, so from the get-go, I said we have to figure out how to shoot as much as we can 'real' or get 'real' elements because we have such a limited time to make the visual effects and to make them look real would be really tricky. So everything in the plan was to follow that brief, and we had to map out when we shoot what just to [keep it all straight]. There wasn't a huge amount of CG effects, but a lot of it came either through small elements or [in combination with special effects]."
In fact, Haug adds that the big revelation for him was how coordinated special and visual effects are for Bond. "Chris Corbould [the special effects coordinator] and I linked up the day I prodded him off of The Dark Knight long enough to sit down and have lunch. We immediately understood what each other was doing and why we were going to work that way. We got together with the First AD [Michael Lerman] and we frontloaded the stuff that was heavy visual effects rendering in the beginning part of the schedule and the stuff that was heavy special effects-oriented at the back end so he had time to build his rigs and then they were going to happen mostly in camera, and we had time to deal with our stuff after the plates had been generated. So we figured out how to schedule it in such a way so neither one of us got too hammered."
Haug cites the two most difficult sequences as perfect examples: the aerial encounter with Bond and accomplice Camille in a DC-3 and the baddies trying to shoot them down in a Marchetti (handled by Double Negative) and the explosive climax in the desert (handled by MPC). "That set of the ESO hotel with the explosions and the DC-3 rig that Chris had to build: both of those had to obviously happen far back in the schedule from his point of view. And that left us to do the skydiving and things we had to do in the beginning. As it is, I still wish I had a few more weeks on some of the CG planes. I think that the lines are heavily blurred as to what's all-CG and what's not. I don't think there's a single thing in the movie that's entirely synthetic."
The aerial sequence was especially daunting. "Most of the successful stuff that we shot were often not in the same shot together," Haug continues, "so you pick the best plate, whichever plane we liked the best: a DC-3 but not always, then you'd go find what you wanted the Marchetti to do in another plate someplace. So we would find the right action, do a postvis, comping the two together, saying this is how we wanted it to work: this one shot on a 50mm, this one shot on a 400mm and then we would take the stuff on the 400mm and map it onto a CG Marchetti and put it into the 50mm shot. Nvizage prevised the whole aerial sequence and [Second Unit Director] Dan [Bradley] was trying desperately to get a sort of map for the pilots so it would make sense what they needed to do, and so it was a complicated process. Dan was not a big fan of previs after having it stuffed down his throat before. And so the fact that he was in charge of it helped him understand what it could do for him. He did something rather interesting: After doing all those QuickTimes, he pulled out still frames from the animations and turned them into storyboards that he handed to the pilots."
All of the interiors in the DC-3 were shot against bluescreen in a gimble at Pinewood Studios. The background was made up of digitally-generated environments, which were mostly made up from plates shot in Mexico. Alex Wuttke supervised the shoot in Mexico for a month, recording data from the shoot and taking detailed reference photo reference for the digital environments. Back at Double Negative the team, led by Environments Supervisor, Guy Williams, used proprietary software Stig and dnDoubleVision to create these environments. A reconstructed DC-3 fuselage was rotated on two axles to roll from side to side and allow the nose to go up and down. This helped to give plausibility to the shots were Bond tries to gain altitude as quickly as possible. The DC-3 also had an engine in it that gave it quite an authentic feeling of engine shake, while onset lighting was pretty substantial to look like it was light from the sun.
At Double Negative, it was also decided to do a very detailed DC-3 model to cover all eventualities and this was modeled by Joel Prager, Jon Veal, Ged Wright, Charles Varenne and Dan Kripac. In the movie the DC3 is seen in extreme close up and brushed metal or aluminum are notoriously difficult materials to simulate or replicate. Led by CG Supervisor Kripac the team worked on extending Double Negative’s shaders to deal with this. In addition, the environment was digitally enhanced and in many cases completely replaced. For safety reasons the stunt plane was not permitted to fly too low, but the environment needed to tell the story of how the Marchetti was trapped, to help convey this the original plates was "canyonized" or extended to make them appear deeper and more treacherous than they were.
Meanwhile, Bond and Camille are forced to jump from the plane with only a single parachute between them. During their freefall, Bond eventually clings to Camille and manages to open the chute before crashing in a large sinkhole. It was decided that a sophisticated combination of CG and live action would provide a more visually realistic approach to the "Bodyflight" sequence.
While Wuttke was on set in Mexico, fellow Double Negative VFX Supervisor Wright oversaw the Bodyflight shoot in Bedford at a commercial facility with a flight simulator. Craig and Olga Kurylenko performed in a vertical wind tunnel in which the airflow is strong enough to support a person above the floor. The actors appear to be skydiving but did not need to wear a parachute during filming.
This method became known as "event capture" and involved shooting the action using 16 in-sync cameras: 8 4K Dalsas, 7 HD Cine Altas and 1 Arri 3 (hand-held in the simulator with the actors). Thus the team was able to reconstruct any digital move they wanted after the shoot using all the cameras. A procedural image-based modeling method was then devised that created a closed mesh representing the surface of the actors at each frame of the event capture.
"That was very dangerous," Forster admits. "And that was a real challenge for Kevin to finish up in time. And that hasn't been done before because usually they shoot parachute sequences against greenscreen with a fan, but your facial expressions will never be the same, and in that chamber it [approximates] the exact pressure of being up in the air as well and gives you the right facial expressions, but still you have to work out the background, which is really hard. This was a sequence I wish I had a little more time to work on."
Once they had shot all the work they used a process called volume carving, which is a hybrid of a lot of techniques: like a pastry cutter, it cuts out everywhere that it's not. A matte of the actor's silhouette for each camera view is projected though a voxel grid that encloses the region occupied by the actors. The value of the projected matte is assigned to each voxel, such that a positive value indicates that the voxel is within the silhouette and a negative value is outside the silhouette. This process is repeated over all cameras, where the minimum projected matte value is persistent for each voxel. Ultimately a mesh is generated from the grid using the zero point level set which represents the surface geometry of the actors, and with projected textures the performance can be used to composite into the final shot.
On several occasions Forster wanted to go outside of the captured areas; fortunately the team, led by CG Supervisor Kripac, also took the traditional route of cloth and hair simulations and body tracking in parallel. Double Negative's proprietary, hair system, dnHair needed to be extended to deal with the velocity of wind that the characters would be exposed to and avoid the hair looking too much like it was crisp, clean and freshly blow dried. The team was able to use digital doubles to realize this and the footage shot in those cases was used to capture motion and apply to the digi-doubles and body tracking was used to match the whole of the sky diving sequence with digital clouds, cloth.
The fiery fight at the ESO provided its own set of challenges for MPC (under the vfx supervision of Angela Barson). MPC had to destroy a hotel with a series of explosions. The external plates of the hotel were shot on location in Paranal Chile while the internal shots were filmed at Pinewood. A small section of the hotel façade was built at Pinewood to be destroyed and used for close up shots.
All of the wide external plates were shot clean in Paranal and MPC did the full destruction in post. 2D elements of fire, smoke, embers and explosions were incorporated with digital matte paintings and 3D debris to give the final look.
The interior shots that were filmed at Pinewood had a lot of practical special effects fire and explosions. MPC had to enhance these shots to make the actors look more in danger. The work consisted of adding additional fire, debris and explosions to the practical on set special effects, as well as making the ceilings and walls look charred and broken.
The MPC matchmove crew built basic models of the interior and exterior of the buildings as well as basic models of the main actors. These were used to enable the CG debris and breaking glass to properly interact with the set and the actors. General Medrano’s room sequence was shot using interactive lighting, but little real fire or smoke. All of these elements, including scorching, fire, smoke, ash, sparks and heat distortion, were added in at MPC. Steam was also added coming from Bond and Camille’s clothing.
"The observatory is a beautiful location, but it was very fragile and they wouldn't let us do much with it," Haug explains. "We built this big four-story set for them to climb out of and all the explosions were full scale and just placed back onto the hotel itself. MPC used a little bit of Flowline [from Scanline] for flying debris, but in the end, most of it were elements we actually shot of smoke, of fire. There was intense pressure to figure out how to get through it so fast because they came in really late. It was the last thing we shot, everything was coming to a head, and that sequence, frankly, was the last one to get written. It got put at the end because it had to be. It was really meant to be a big, practical flame. And what we mostly did was rev up the intensity when it fell off during the parts of the shot that we were using. At the beginning of the shot, they were generally always great, but then the sprinklers go off and they have a tendency to dampen everything down. And so we'd bring it back up to the original intensity. And there were a couple places where they got really close to the fire but not close enough to look good, so we brought it a little closer. And there's the final sequence when they're trapped inside, where, again, the biggest problem was making them feel really trapped. There has to be proof of that so we brought the flame within inches of them just to make sure there was no doubt in anyone's mind."
In terms of the water element, the boat sequence (handled by Machine FX and supervised by John Lockwood and Steve Street) turned out to be easier than expected, according to Haug. "The biggest shot is the one where he rams General Medrano's boat, done in multiple parts. I think it was the first shot that Dan ever talked to me about. The initial plate was shot in Panama and the jet boat that comes in and flies around Daniel was a tricky one because when you're doing boat to boat like this, it's pretty uncontrollable. We got one that was good and then we matched that angle with another piece attached to a boat that actually rams Medrano's boat and then we matched that same angle with a stunt on dry land with the guy getting yanked off to the end of the rope and falling to the end of the boat. We put it on a slider so it looked like it had impact. And then Machine just took those pieces and put them all together."
The boat sequence also included some CG shots where Bond on a motorbike crashes through loads of CG wooden packing crates, and then lands on a CG boat deck, which Machine rebuilt to make the stunt look more dangerous. He then jumps over onto a second boat, which Machine also rebuilt again to add to the stunt. It was also a big rig removal and cleanup effort, and a head replacement for the stuntman. There were other complex comps with CG enhancement such as boats crashing into one another, throwing Bond into the boat, which required matching up three different plates of live action, bluescreen boat and stuntman being hit and warping the various elements to tie it all together and make it seamless. CG objects were added to the boat to tie in continuity and make the shot a bit more dynamic. There were many other head/face replacements in this sequence, as well as bullet hits, debris, etc. Machine used a mixture of off-the-shelf tools (Shake, Maya and RenderMan for Maya, Matchmover Pro, Photoshop and Combustion) that have been customized here, plus some in-house scripts.
Meanwhile, the early Siena foot chase, in which Bond pursues an MI6 traitor, which culminates with both of them falling through the roof of a gallery and then fighting on scaffolding, was handled by Framestore (supervised by Jon Thum). This turned out to be more complicated than expected.
"The floor shot where they crash was supposed to be so simple," Haug admits. "It was 90% stunt and 10% help. And it turned into this monster. There were four parts to it and you're only looking at three of them. The final part, where the stunt man was to turn into Daniel, didn't end up happening because the stuntman ended up falling off the scaffolding and just managed to catch himself, and Marc liked it so much that we changed the whole choreography to stick with that. So the fourth part with Daniel got dumped, beginning with him falling. Daniel's willing to get on a rig and do anything. The only thing he didn't do in that sequence was actually going through the glass. That had more to do with scheduling. I think he was already off in another country when we did it. We learned from previs that the art gallery set, which filled the 007 Stage at Pinewood, wasn't big enough. Marc kept saying that it didn't work for him: it was too close to the ground, they should just slide down the rope, jump off and run out of the building. Then we would have to continue the sequence back out on the streets of Siena again, and nobody wanted that. So we ended up having to make it 50 feet taller, so that when you were all the way on the bottom you still had a long way to go. And clearly you weren't going to make it, so you didn't want to let go. We put it on the side of a hill so it looked taller when you approached it."
Finally, the opening, pre-credit car chase in Lake Garda, Italy, was handled by Double Negative as well. Wuttke supervised the shoot, which covered all the exteriors and some practical interiors with a stunt driver. The chase travels the ground from Lake Guarda to its culmination at the Carrera marble quarry. The majority of the exterior work was filmed from remote camera positions and in up to half the material there were many fixes required, including removing up to five or six cameras, crew members, a stunt rig, while also adding damage to the Aston Martin and a face replacement on the stunt driver. Windscreens were taken out for the shoot, so these were later replaced with CG windows, which needed appropriate reflections and also CG shattered and damaged windows where appropriate in the sequence. The cut changed regularly during the editing, which meant that the team was often called on to remove cars at certain times and add dirt and dust to other cars as they were edited in later in the sequence.
Practical interiors also included replacing the rear view mirror with bluescreen elements of Craig, who was not available during the shoot. These elements were all shot at Pinewood where the Aston Martin, the Alfa Romeo and a large truck included in the action were mounted on a big trolley and shot against bluescreen. Due to the enormous stage required there were many areas were the bluescreen was not applied and a great deal of roto work was needed to make the shots work in the composites.
It was critical to get the camera height and framing all exactly right or it would be impossible to make the shot look real. The interior was an extremely technical shoot and there were many variables -- the set lighting was set to match the plates as closely as possible -- but even so, a big part of the post job was to re-light the shots in order to assist the integration.
"It's an odd way to work," Haug admits, "usually you shoot foreground first to get the right angle on the actor and you let the actor drive the background. In this case, the background drove the shots of the actor. Luckily, in a car shoot that's pretty easy to know that's going to be a good angle."
Bill Desowitz is the senior editor of VFXWorld and AWN.