<em>Speed Racer</em>: Go Go Go Anime

[1]

Speed Racer looks to anime to break new ground in digital cinema. The film contains more than 2,000 vfx shots, including this thrilling racing sequence that shows car flips [1]. All images courtesy of Warner Bros.

A lot has changed since the Wachowskis redefined the cultural landscape with The Matrix [2] back in 1999: raising the stylistic bar with their imaginative brand of next-gen, futuristic retro storytelling. Movies and videogames have been trying to catch up and surpass it ever since. But after a string of stunning cinematic successors, ranging from Sky Captain and the World of Tomorrow [3], Sin City [4] and 300 [5], the Wachowskis and their collaborators have now taken next-gen to the next level with Speed Racer (May 9 from Warner Bros.).

Once again, the Wachowskis look to the past -- in this case anime -- to break new ground with digital cinema, which is where the future of movies is headed. And in Speed Racer, which contains more than 2,000 vfx shots from nearly a dozen studios, they have literally found the perfect vehicle to explore the power of idealistic virtue and artistic freedom through the essence of anime. From the opening images of racing car drawings in a flipbook, the metaphor of Eadweard Muybridge's [6] objects in motion is conveyed through a number of references right on through "Bullet Time" and beyond (with zebras, no less). But with Speed Racer, we have a whole new convergence of escapist interactive thrills where anything is possible.

"They were in a frame of mind to re-imagine something that influenced them as young people," recalls John Gaeta [7], who, with Dan Glass [8], spearheaded the internal vfx production effort with many of their previous collaborators from The Matrix trilogy. "As we now know, they are appreciators of anime. And like late baby boomers or Gen-Xers, Speed Racer was one of the first animes to land and grab people's attention [back in the '60s]. Really, when you look back at that cartoon and its production values, it's very simplistic animation -- more so than most. But you don't remember it that way. It reminds me of Narnia [9] in that it has a different energy and magic reading it as an adult as opposed to reading it as a child. If you do look at the story threads, they are quite edgy, and there's a lot going on concerning the relationships and obstacles in terms of the family against all purveyors of fixed racing and corporate espionage. The first conversation had to do with the story of the family against all odds, and then a good action subtext for them to glitterize and modernize and Wachowskiize in the way that they would love to see cars today."

Gaeta, like the Wachowskis, is a pop culture sponge and a visual innovator, who often talks in code as a way of conveying design ideas: "Poptimistic," "Techno Color," "Photo Anime," "Faux Lens," "Virtual Cinema," "Car Fu." In fact, he describes Speed Racer as "Sin City mashed against Matrix techniques pushed through a Pop Art, live-action anime format and wrapped in a genuinely cool narrative."

On behalf of the directors, Gaeta initially oversaw the goals, design and content objectives for the Speed Racer previs team, while in parallel Glass pushed into the innovative HD production pipeline (arguably the most advanced ever). "We also talked about our frames of mind collectively today with regard to cinema and format and texture. Speed Racer expands on what they began with The Matrix trilogy in terms of 2.5D moviemaking and even what we've seen with such next-gen successors as Sky Captain and the World of Tomorrow, Sin City and 300, which hadn't come out yet [when we started Speed Racer]. But Sin City caught my eye in its attitude and approach toward the realization of a graphic novel, which I thought was a very free creative approach that [Robert] Rodriguez had taken. I definitely thought that he was actually on to something with regard to [presenting a hyper-reality]. But you'll find in interviews that Rodriguez was influenced by the first Matrix to some degree. Sin City was approached from a very storyboard heavy, comic book style form of photography. And, of course, The Matrix was derived from Japanese anime.

[10]

The filmmakers discussed how stylized and hyper-real to go with the film. The goal was to remake the components with spherical photo elements and super saturated colors to achieve "Photo Anime."  [10]

"There were lots of choices on Speed Racer about how stylized and hyper-real to go. There were lots of ideas left on the table from The Matrix… but there were lots of possibilities in deepening the form, which weren't as doable as they are now with the abilities of digital cinema end to end post processing. So we said, let's loosen up and move away from this deep photorealistic and integrationist approach, and try to pursue the nuances of anime a little closer to the bone. And so that began this slow but steady remaking of the rules of this universe. It seemed that to pay homage to the spirit of this animation format, we needed to find the language that the average viewer can understand. Camera movement and editing and in anime when there are rays behind an object for graphic impact, they know what it means. It boosts the energy and emotional underscoring of an event. We deconstructed a number of anime to find the basic language and started to translate that graphically into our conception of what the movie should be. Our goal was to pursue what I call 'Photo Anime' to remake the components with photo elements. Rather retro in some of the approaches, but re-thought in as much of a modern digital pipeline as we could muster. So we decided to take a look at what's going to drive the decision for digital cinema. We didn't want something tune shaded like A Scanner Darkly [11], but we wanted to feel the qualities like pure color and definition and an obvious understanding of layer to layer because animation tends to have parallax emphasized because it is actually constructed as layers that are moved in crafty ways to give you the perception of a camera moving through space when it is really just a down shooter or a rostrum camera. And so we endeavored to make a virtual rostrum camera and replace our elements.

"We decided, of course, to shoot our characters with HD cameras. We tested a lot of cameras with [DP] David Tattersall, but we got inside information that Sony was about to put out a camera called the F23, which was the next-generation of the CineAlta. It turned out to be a superb camera and was far more fitting than other cameras in that it was a 3-chip RGB and 2/3" chip, which allowed us to have superb depth of field that other filmmakers are working hard to get rid of because they want everything to look cinematic with cinema lenses, whereas what we wanted was the option to have infinite depth of field and customized lenses at any time. We wanted to combine a telephoto foreground with a wide angle background. Stylistically it was cool, so that was a major reason why we went digital. Also, we wanted a grain free, noise free and essentially texture less-looking end result with extremely vivid colors, which at times we called 'Techno Color.' And so, in that pursuit, we realized that we wanted many things more than film was going to offer us. We dreamt about the day that the film moved beyond theaters and hoped that digital cinemas would be more prolific at the time. But we were thinking about legacy media, including Blu-ray and HD broadcast, so we really wanted the film to look pristine in, say, a Blu-ray format. What we wanted is a digital master all the way through that's clean and hyper pop or optimistic colors, which we call 'Poptimistic,' and collaged colors and contrary colors to emphasize depth and energy. In our construction, what I would regard as an umbrella effect turned out to be the making of this Poptimistic live-action/anime format, and while you can look at any novelties we've conjured for any individual shot, it's really the hope that the tone and feel of the movie would be a different type of experience. We wanted to provide that to Larry and Andy in helping step them through the process of how to approach production design and how to skate through the end frame of all frames in the movie, which was quite an endeavor.

[12]

The characters [12] were shot with HD cameras. Luckily, the filmmakers got their hands on the new Sony F23, which allowed for a superb depth of field to give a feeling of infinity to the backgrounds in the film.

"Going after this anime and finding the essence of the form and texture with some form of hyper-reality led us to some of these decisions. That's the DNA. What it allows us to do is make rules for how we approach a shot, how we compose and composite in post. It's virtual cinematography, it's done with faux lenses, there's a degree of what we call editography, a collage-like editorial form created by Larry and Andy that is expository and delivers a lot of information in a fluid way that is not the same as an un-breaking shot. Again, it is borrowing from the way manga and anime tell stories. One anime that we looked at a lot was [Hayao] Miyazaki's [13] The Castle of Cagliostro. He found elegant and expressive ways of allowing events shown and characters to be detailed. In Castle, he takes often optimistic colors to the maximum and allusions to how people and vehicles move through spaces."

Photo Anime crucially entailed revisiting the 2.5D layering method of placing both foreground and background elements in focus. "Again, a throwback to the way traditionally painted or drawn backgrounds used to be done in animation, but essentially still frames that the camera scrolls across," adds Glass." And we wanted to replicate that in a photographic form, so we jumped on technologies that we've used for many years now for our virtual backgrounds and we actually made them a little more primitive. In fact, you can say that we've regressed on this. What I mean by [that] is we took away a large amount of the geometry, so we're literally taking a still, a plane or, in most cases, a spherical panorama. We would shoot it ultra high-resolution, bracketed exposures as high dynamic range and build that as a spherical texture. And then composite a scene into that sphere, so that when the camera panned and tilted, it would obviously retain the 'bubble' accordingly. But even as you track, zoom and dolly in toward characters, we would imitate that with the background by scaling it. And although it's not technically correct, emotionally it works and your brain understands it in a similar way that animation works: you get the sense of movement even though it's not totally true to reality."

The production, which was based in Germany, had its own in-house stills group called the World Team, headed by Lubo Histrov (visual effects environments art director), whose company, Christov Effects and Design, executed most of the key spherical bubbles in 12K resolution along with creating digital matte paintings and art concepts. Some of the racing tracks had to be made in CG while others were made with photo components taken from around the world: Morocco, Greece, Italy, the Alps and Austria, hence the title. "We scoured that region of the world with multiple location crews and scouts, trying to determine sweet spots to send our stills team," Gaeta continues. "In all my years in filmmaking, I've never had a team like this. We had the highest skilled, specialized photography unit, one that was comprised of people that usually would be located in vendors and the ones that we would rely upon to do spherical, 360-degree captures of all sorts of other textures in order to do image-based environments. So we built this team and they shot thousands of 'bubbles,' as we would call them, and going into certain places with the idea of how we might construct and gang these bubbles up: recombine them in strange ways and make these Photo Anime backdrops. So the World Team was doing that in parallel through pre and during photography while we were planning the action independent of certain environments.

"The car choreography involved a lot of combat as well as stunt-like scenes. But, at the same time, the World Team needed to have every single backdrop photographed, particularly ones behind actors, and delivered to another lengthy process, which involved enhancement to the captured locations. We'd do QTVR dailies and they'd select locations that we'd then pass on to conceptual visualization people, including creative composite people like Steven Lawes and matte painting people like Lubo Histrov, who would help figure out how we were going to slice and dice these bubbles and then also color enhance them in the preferred style at very high-resolution. The goal was to be there on set for every single greenscreen shot with the HD navigatible background that could be comped while shooting. This allowed [the directors] to strategize with DP David Tattersall about how light and color on set would marry against these fantasy environments. Things were semi-integrated. Lots of stylistic lighting decisions had to be made. In order to have a tight bond with David, Production Designer Owen Paterson and Costume Designer Kym Barrett, we devised a method to having all of the material and the navigatable versions of it available. Digital Domain built a new software called Sparky [named after the Racer sidekick] that allowed us to composite on stage in HD while shooting every scene and composing the backgrounds and doing weird lensing options, while visualizing 30-plus minutes of psychotic car racing scenes (Formula One mashed with X Games and videogames), which also had to be passed through this Photo Anime layering."

[14]

The car choreography involved a lot of combat as well as stunt-like scenes, called "Car Fu."  [14]

According to Digital Domain's Kim Libreri [15], who heads the new technology efforts, Sparky was a necessity because of all the photographic bubbles that had to be integrated. "We wanted to make sure that David [Tattersall] had as good a feedback [in realtime] to know what these characters were going to go into while he was lighting and just to make sure that the key to fill ratios matched the backgrounds and colors. What would happen is that Dennis Martin would go out and shoot them, Lubo would paint them and that would be the bubble for the character. So what we did was take a games-level Intel PC, popped an NVIDIA 8800 GPU into it and then bought a Blackmagic DeckLink high-def capture card, put the whole thing together, our engineer Brian Smith spent six weeks writing software, and that was Sparky."

The CG vehicle design and racing infrastructure were also put together by Digital Domain, along with a couple of the major track races: the hometown Thunderhead contests and the climactic Grand Prix; CafeFX did the Fuji Lexicon race and a combination of BUF and Sony Pictures Imageworks executed three major sequences, including the cross country rally called the Casa Cristo 5000. All of actual racing action took place in a small cockpit that sat on a motion base and then they either uploaded files into the motion base or played live with the gimbal operator moving the gimbal per instructions by the Wachowskis. The CG car shells went through an extensive design phase for more than a year. There were dozens of custom cars but came up with a concept to have track racing vehicles rotate all four wheels 180 degrees in either direction, which they called the "T 180," so effectively they could do a four-wheel rotation. That enabled the cars to get into choreographed martial arts action, or Car Fu. And they threw in the jump jacks from the TV series, which could be controlled independently, and they could perform aerial acrobatics in these vehicles, speeding at 200-300 mph. The tracks were closer to a skateboard park than a Formula One circuit.

"We very much used previs as the first stage of animation set up," Glass adds, "and in many cases getting us most of the way down the animation pipe. We did a lot of conceptual tests for how the other material would work in terms of the way layers might move and shift behind people.

"We used a lot of off-the-shelf software. We knew we were going to use many vendors and wanted the previs scenes to be as useful as possible. The basis was Maya [16], and, although the rigs could be considered proprietary [they were a custom component that worked in Maya], Digital Domain allowed us to implement them in our scenes and also to offer them to other vendors. Digital Domain rendered in mental ray and did a lot of work building custom shaders in collaboration with mental images for the cars and the paint work to produce stunning-looking vehicles.

"On the composite side, Digital Domain and some old colleagues of ours built some custom kernels for doing focus filters, which became quite a motif throughout the film. One of our initial observations was that earlier animation sometimes has a primitive defocus filter and everything has the same level of defocus because it is a flat plane beyond the foreground characters. If you seriously crank the defocus level, it really pops the characters from their backgrounds. Introduce different shapes -- a circle or a hexagon or the shape of an iris and diamonds and rectangles and chevrons and hearts -- and the emotion or message in a scene was sometimes reflected in the way the image defocuses. The particle pipeline was Houdini [17] based in conjunction with Maya and rendered through mental ray.

[18]

The digital team worked on creating on-set systems that allowed the actors to drive in motion bases so they would feel like they were in real cars. CG cars and environments were then placed around them. [18]

"The experience has been very freeing. The all-digital pipeline is very, very close and I think this movie was one of the first where the transition was relatively smooth. This felt like the first time the technology didn't hinder in any way what we were trying to do and enabled us to have a great deal of creative play with the images. That to me was one of the most exciting aspects of it."

Look development on the visual style and car design originated with Gaeta, Glass and Libreri and Mohen Leo (two other Matrix alumni) at Industrial Light & Magic before Libreri and Leo moved over to Digital Domain. Leo worked on setting up the systems on set that allowed the actors to drive in motion bases to permit them to feel like they were in real cars and then would have CG cars placed around them; they weren't just Digital Domain's shots, but also other vendor shots too. But they set up all the technology for that. Digital Domain also set up the animation previs pipeline for the cars so that the previs team could work with cars that would ultimately be transferrable into post.

"We did proof of concept animatic test for designing tracks from the normal to the super wacky inspired by modern videogame and anime aesthetics as far as loop-d-loops and other spirals, hump backs, spiders, massive drops, cars falling off the tracks and doing aerial combat," says Libreri. The production inevitably coined the term "Conveyor Belts" to describe the infinite loops of road to put CG cars on. Overall, Digital Domain worked on 445 shots, 300 of which were fully CG.

"In Berlin, the art department had around 20-30 cockpits that were mounted on a hydraulic motion base in front of greenscreens and a number of systems there to allow the directors to view the cause and context of the backgrounds," Leo explains. "Encodacam was there with a system and Brainstorm so that when the actors were sitting on the motion base in front of a greenscreen, the directors could frame the shot and get a live preview of the cars around the actors. And we had a variety of systems to steer the motion base to give the actors the sense that they were driving and get the correct camera moves using everything from traditional live waldos and joysticks to pre-programmed moves on the Cooper system to a high-end car simulation racing engine called rFactor from Image Space. It's a very open system and we worked together with them to create an interface between rFactor and the Cooper System at the motion base. We had the ability to actually place a car on one of the tracks. For example, we built Thunderhead as a full 3D track in rFactor so that with a steering wheel and pedals behind a gaming system [you could] drive around the track and then record the accelerations and the motions of the car out of the simulator and feed that onto the motion base to get the appropriate motion like a series of turns around the track. Another thing the simulator allowed us to do was give the directors a gaming pad, which we mapped the buttons to, allowing them to simulate the car being hit by another car or jumping and landing for the car fu action. The simulator was a great help to Second Unit Director James McTeigue, who could see our stunt drivers driving around and give them instructions. The actors tried a live test from the cockpit that allowed them to drive while viewing the track from the simulator on the monitor, but it was too difficult for them to act and essentially play a videogame at the same time."

Digital Domain additionally set up a new mental ray rendering pipeline. "The amount of technology that went into our global illumination [was immense]," Libreri continues. "Using traditional methods to render our cars [19] using mental ray would take about 10 hours a frame; they got our renders down to about 10 minutes a frame for cars simply by doing massive optimizations to the global illumination code for mental ray. Maya was our front end for cars and animation and rendering and Houdini was used for effects animation. Both mental images and Autodesk helped us take the Maya integration to a new level. And Atomic is our new lighting engine that fits on top of Maya that acts as a layer and light manager."

[20]

Digital Domain additionally set up a new mental ray rendering pipeline. Traditional rendering methods for the cars would have taken 10 hours a frame.  [20]

Meanwhile, Paris-based BUF Compagnie not only contributed to both days of the Casa Cristo rally (including starting line, arches, dunes and cortega finish along with the ice cave and Brandenburg finish), but also created the main city, Cosmopolis. This was fully recreated in 3D but shown through 2D processes. BUF was also responsible for the interiors of the futuristic-looking Royalton Industries building (including the tram ride, bizarre training labs for pilots and the vertical assembly line for the car factory). This was all under the leadership of CEO Pierre Buffin, VFX Supervisor Geoffrey Niquet and VFX Producer Simon Vanesse.

The studio's renowned camera mapping technique was used for the reconstructed environments such as the arches during the rally race. BUF's entire 3D pipeline relies on proprietary tools developed by its in-house R&D team. Post-production is also done on homegrown software at each stage, from modeling to rendering, including mapping.Specifically for Speed Racer, BUF developed a program that allows for 3D camera movements with 2D treatments. Also, animation of the cars was a mix between keyframe animation and simulation.

Santa Maria-based CafeFX, which handled the Fuji Lexicon race, had to grab so many people so quickly that it worked out of the Santa Monica office of sister division The Syndicate. "There was not a lot of time for experimentation," relates vfx supervisor John Van Vliet. "I always tell everyone that the most exhilarating and terrifying thing in a motion picture is a complete blank page, and this was certainly one of those moments. They gave us a fairly complete animatic, a book of sketches and some photos, which were their inspirations. This was originally a Digital Domain sequence, but they were pretty much up against it. They did some preliminary layouts and modeling work on the cars and we had to flesh it out and figure out what the actual construction of the track was and had to build the track. A big hurdle was road graphics. They had a very colorful concept but my concern was that it was so colorful that it would camouflage the color of the cars. And the other thing was that the way the graphics were created, as the cars went over and the camera would be on them, that there would be strobing and I was afraid of the reverse wagon wheeling effect.

"We took inspiration from some of their concept art and stretched it way, way out like a Polynesian tribal tattoo, which was in line with what their thinking was. And they went for it. So when the camera gets close to the road surface, it actually animates and snakes out and we were quite proud of that. Our concerns turned out to be correct with colors, and the cars disappeared in a lot of the earlier stuff, so our solution was to go back to the earlier part of the race and change the graphics to a dark blue and after we eliminate the cars we go back to a more colorful orange and red. Our main compositing tool was Fusion. And we also had to work in explosion work that had an animated feel. When we got into pyro stuff, they actually talked about doing old anime, so we brought in a guy with anime experience, Jeff Howard, who's done this, and we made beautiful anime explosions. But then they changed their minds and went back to doing it in CG in 3ds Max [21]."

San Francisco-based Evil Eye Pictures was tasked with handling the opening school flashback, which provides the first glimpse of the exterior Speed Racer universe. The boutique was did an anime-inspired go-cart explosion during another flashback sequence with Speed and Rex in their garage.

However, the primary work consisted of the Rendezvous fight sequence shot on greenscreen. The World Team shot locations in a mountain pass, which were turned into matte paintings and then bubbles. "We were handed that and modified it quite a bit to use as main background," explains Exec Producer and Co-Owner John Jack, who collaborated with VFX Supervisor and Co-Owner Dan Rosen and VFX Supervisor and Co-Owner Matt McDonald. They had previously worked on key fights in The Matrix sequels, so this was a perfect fit. "We added CG snow and other graphics to create a power feeling through punches and kicks. A punch would streak through the snow and displace the snow as it went through it and leave a trail. That was a combination of 2D and 3D work. There was a lot of design work involved in stringing multiple greenscreen plates together because they wanted long, flowing shots that went from one section of the fight to another. We strung together eight or 10 different plates for some shots of individual people fighting and we basically zoom or snap pan or push through from plate to plate.

[22]

The main challenge of the Rendezvous fight sequence [22] was creating the graphic design of the snow.

"The main design challenge was the graphic design of the snow. We ended up doing a snow library using Maya and mental ray to render that out into over scan passes that we could layer up in Z, but given the number of shots we didn't want to do full 3D snow for every single shot. We had to do a 3D track and body match moves for all the snow interactions, but we basically did multiple levels of snow. The other issue was they wanted snow at the beginning of the sequence to be lightly falling and then increase throughout the sequence until the crescendo of the fight where it comes down the heaviest. We did make custom 3D falling snow for severe down angles. We were originally asked to do Hallmark-looking snow, but that graphic design was modified."

For Speed Racer, Evil Eye used Shake with many custom plug-ins, macros, Furnace, Maya, mental ray, Silhouette, Combustion, PFTrack, Syntheyes and Photoshop. The company has a custom proprietary plug-in for shake that allowed them to use 360-degree spherical environments directly within Shake.

[23]

An amazing shader was used to design the ground of the desert flats, which changed color and texture. Another nod to anime was the effect of leaving incredibly long dust trails.  [23]

For VFX Supervisor Kevin Mack of Sony Pictures Imageworks, Speed Racer offered an especially fulfilling artistic change of pace. "When they came to see us, they said they thought of me because of the work that I had done that was artistic or stylized [What Dreams May Come, Fight Club [24], Big Fish [25]] whereas a lot of the folks that they had worked with were having a hard time with it because they were so accustomed to doing photoreal, and they really wanted the film to look like moving Pop Art. We lost our first month getting data from other vendors and models having to be re-rigged and texture coordinates. Our sequences were self-contained and we had an idea of how we wanted them to look.

"There wasn't much in the way of design for the desert flats, so we put together an amazing shader for the ground, which changed quite a bit from yellow/gold to red/orange, and from the tile cracked mud to little strings of salt flats to gravely lava rock. So basically on the ground you have this range from yellow to red/orange. And so I do abstract art and am into some of these notions about the perceptual effect of color, so it seemed like a good opportunity to use this super complement, which is a thing I've come up with, which is that blue and orange are a very powerful complement, I believe, because it's what we've woken up to since we've crawled out of freezing caves to the warmth of the sky. What I found is you even get a stronger effect if you get a range of complements. For instance, if you have a grad of yellow to red/orange, then you have another grad,which has a range of colors from turquoise to blue/purple, you've got 2½ complements working at the same time. It's so stunning and then you have the colors of the cars and all the highlights and the sun.

"And the other thing they wanted to do was find little tricks that harken back to anime. And so we have this effect of leaving incredibly long dust trails. So the idea was to recreate the hard edged, beautifully animated look of cloud forms and so I thought it would be cool to apply a fluid simulation and get a very realistic, very dense cloud of dust being kicked up by the cars, and, of course, as they're flipping over and doing all of their car fu, that they would be passing through these things and affecting them and stirring them up more. To me, it seemed like the idea was to get as much realism as possible at a certain level so you could depart from it even further, like the saturation of the colors. Also, with the fluid sim, we could treat it in the comp in terms of the colors, density and contrast to be a little more like that hard-edged, ink and paint anime approach to smoke. They only specified a look up table for them to reference. We did a multi-plane exaggeration through the manipulation of the bubbles."

For John Knoll [26] of Industrial Light & Magic, Speed Racer was a little outside of his normal comfort level, so he really didn't mind only playing a minor role after Libreri and Leo moved over to Digital Domain. Knoll supervised a single action sequence involving Racer X and the Cruncher Block truck.

"We used assets that mostly only played in our sequence with the exception of the Racer X car from Digital Domain," Knoll recalls. But the truck and environments only had to work within our sequence. It was one of the few nighttime scenes on a lonely highway as we are introduced to Racer X in his encounter with the Cruncher Block truck. There's all this talk about going for a hyper-real anime look. It's not supposed to look like photography. But it's hard to define what that really means. There was a little concern going into it about an unlimited space to explore, and you've got to get it done for a deadline. But it actually worked out pretty smoothly. It meant that a lot of the instincts that I usually rely on when I'm looking at shots and deciding where I want to go with them didn't really apply. Things like: if you were really shooting this, what kind of camera mount would it be? What that mean about the style of the framing? What would the vibration be on this? With the helicopter move, they're probably on a long lens and probably can't get too close to it. You're going to drift around a little bit. Also, photographically, if this was two vehicles on a lonely highway at night, it would pretty much be entirely lit by headlights. But if you were doing anime, you would draw it so you could still see things.

"I just tried to get into the spirit of what anime is. But you still have to tie it to some live action because we had tapes of characters in the cockpits of their vehicles, and had to match the photography. The general problem was there would be a realism gradient, for lack of a better term, where things would be photographed right around the camera, and, as you got farther off into the distance, it's OK to be more and more stylized. For example, on this lonely highway that we are traveling down, the trees are all textures on cards. They're pretty much the same tree over and over again. Normally, you'd try to mix it up and give it a nice organic look, but we kept being told to go more for that 'repeato tree thing.' So we made them a little more regularly spaced. In anime, you're always fighting with how much work it is to do a completely organic environment like that, so you end up with repeated cycles. We were trying to inject that animated cycle thing into this.

"We kept hearing: 'You don't have to have things move correctly in perspective.' In a side shot you can have strolling layers often done in anime. It is an interesting experiment and was fun to do. And in the end, we came up with a stylized look for our sequence that I thought was pleasing, anyway. I had a very small crew and we were experimenting with a little different workflow from [the assembly line approach that] we normally do. Almost everyone did two or three steps along the pipe and I got some really good feedback from the artists."

In refashioning "sample cinema" methodologies, Gaeta ultimately hopes that Speed Racer will be remembered as a genuine creative exploration and not as much as a technological one. "Dan and I were able to attack any problem from either side of the brain with such a talented team. They were simply the most talented, experienced, well-rounded and fastest I've ever worked with. That was truly the main reason we were able to create an end to end approach, nearly touching every frame in the movie. The visual effect centerpiece was creating the movie itself within [this new Photo Anime] format."

Bill Desowitz is editor of VFXWorld.