Step By Step: Quantum of Solace

Upping the ante for the 22nd James Bond movie was no small feat. But the Columbia Pictures release Quantum of Solace boasts a novel approach to a jaw-dropping sequence in which actors Daniel Craig and Olga Kurylenko leap from a burning plane and plummet to a sinkhole in the desert below. Director Marc Forster and Visual Effects Supervisor Kevin Tod Haug challenged London-based Double Negative (DNeg) to design this 15-shot sequence in a way that would let audiences see the actors up close as they tumbled to earth.

DNeg Visual Effects Supervisor Alex Wuttke says he considers this bodyflight sequence among the most difficult of the 340-plus shots they created for the film. “Kevin wanted to to capture the event entirely in a photographic way and then choose the camera moves later,” he says. DNeg's solution was to embed a semicircular array of 15 synchronized digital cameras — eight Dalsa 4K Origins and seven CineAlta F23s — into the walls of a wind tunnel, where the actors and their stunt doubles were photographed being tossed around. Tracking markers on the tunnel wall helped DNeg calibrate the cameras' relative positions. This event-capture approach yielded about 140 degrees of coverage from the cameras' POVs, and having all this footage allowed DNeg to reconstruct a dynamic sequence in post.

“We came up with this technique by doing research into computational photography, an area of computer graphics that is quite new,” Wuttke says. “Previous approaches, like ‘bullet time'' in The Matrix, wouldn't have worked because we'd have had to be prescriptive of the camera move ahead of time, and we couldn't have choreographed a camera moving around people who were tumbling out of control. In essence, it's taking a stereo pair of cameras with slight offsets between them, so you're looking at images in the computer and finding the parallax between those points — which effectively gives you depth. That was the foundation for the technique.”

“On top of that, we positioned cameras in high and low positions outside of the main array,” says DNeg 2D Supervisor Victor Wade. “That gave us additional unique viewpoints. It helped with the reconstruction to see the actors from as many different angles as possible. We had a huge volume of data from which we made selects.”

Heavy rotoscoping with DNeg's inhouse tool Noodle was necessary to lift the characters off the background. “Then we used a process called ‘shape-carving'' using silhouettes,” says Wuttke,who notes that proprietary software called Double Vision enabled this process. “We used an algorithm that looked at the matte left over from the roto — wherever it appeared solid was inside the subject, and wherever it was black was outside. We carved out the geometry to effectively give us the actors' shapes in 3D space, and we did that for all the different angles that we'd captured from the photography.”

This process yielded human shapes in rough-mesh form, which the DNeg team dubbed ‘garbage bag'' people. “They looked like real people covered by plastic bags,” Wade says. “But we could see the essence of their action — your brain tells you that they're people.”

Working in parallel was a 3D team that built high-resolution digital
doubles that could be tracked onto selected shots as needed. “Because
we had the action captured from all those different angles, we could
track digi doubles' faces on top of this footage in a pretty
straightforward way,” Wuttke says. “We used a mixed bag of [The Pixel
Farm] PF Track, [2d3] boujou, [Science.D.Visions] 3D Equalizer, and an
inhouse tool that uses distance constraints. We can do 2D tracking
within [Apple] Shake and export those 2D tracks into a bit of software,
which turns it into a 3D matchmove. We used a mixed bag.”

The strong winds in the wind tunnel distorted
the actual faces, which had to be matched in the digital doubles that
were animated in Autodesk Maya. “We wrote skin deformation tools to
send ripples across their skin, to match synthetically what we were
seeing in the footage. These digi doubles were complete with hair and
cloth simulations for their clothes. So the net result was that we had
the flexibility to mix bits of real footage with CG,” Wuttke says.

Lighting was extremely challenging, however,
since the sequence begins in a sunlit sky and ends inside a dark
sinkhole — and the faces had to be seen full-frame. DNeg used Pixar
RenderMan to get subsurface scattering effects of light on the skin of
the digital doubles. “Because we had captured the motion so
effectively, we could just track our digi doubles on top of the
photography, relight them, and put maps from the digi doubles back onto
the photography,” Wuttke says. “It was a hybrid solution.”

All of this happens against backgrounds that
were combinations of synthetic elements. “The environment was entirely
CG — including the plane, clouds, and smoke,” Wade says. “We also see
the desert and the sinkhole that they're falling into. All of those
environments had to be modeled, textured, and lit.”

The mix includes a matte-painted sinkhole
created with Adobe Photoshop and desert panoramas stitched together
from digital stills. “It was a huge environmental effort because no
plates were shot,” Wuttke says. “There's a lot of tiled photography. We
have a tool called Stig that essentially takes bracketed photos and
stitches them together in ultra-high resolution, like a cyclorama. We
also projected some photography onto 3D sky domes for moments when we
needed parallax inside the image.”

All of these elements were composited in
DNeg's version of Shake, according to Wade. “We modified a 64-bit
version of Shake inhouse, and we did all our compositing with that,” he
says.

To manufacture such a key sequence for Quantum of Solace,
DNeg had to combine real and synthetic elements to an unprecedented
degree. “It's very satisfying to do something that hasn't been done
before and do it well,” Wuttke says.

Director: Marc Forster

Visual Effects Supervisor: Kevin Tod Haug

Visual Effects Supervisors: Alex Wuttke, Ged Wright

2D Supervisor: Victor Wade

3D Matchmove Supervisor: Joel Prager

Sequence Lead: Brian Kranz

Lead Animator: Chris Sweet

CG Artists: Theo Facey, Chris Kilshaw, Gavin Harrison, Guy Williams, Nick Petit

2D Lead Compositor: Tilman Paulin

Compositors: Mike Bell, Paul Stirling, Nik Brownlee, Dean Koonjul, Patrick Nagle, Isaac Layish