Prevising 'Charlie' for ‘Burtonvision’

Production designer Alex McDowell, known for creating wild and spectacular sets for Fight Club, Minority Report and The Cat in the Hat, was taken a bit by surprise when Tim Burton indicated that he wanted a 60s pop art look for much of Charlie and the Chocolate Factory, referencing, of all things, the Italian space-age cult movie, Danger Diabolik.

And whereas David Fincher or Steven Spielberg will explain very clearly what they want and need, McDowell soon discovered, like so many others before him, that Burton works more instinctively and organically. The director responds to certain images put in front of him or he doesnt, and the challenge is to build from there.

But McDowell, who in the U.S. is accustomed to making full use of 3D design as part of the art department before turning it over to previs, was instrumental in establishing a unique centralized previs unit in the U.K. while working on Charlie in London.

For the Charlie budget, the art department and the vfx department split the cost of previs, since there is a point in production when the bulk of the previs work switches from the live action to the CG planning, McDowell explained. What happens generally, and happened on Charlie, is that the amount set aside in the art budget for previs expired long before the end of the previs term. The previs team continued to work 40 weeks later, until the end of shooting or beyond, because the more we do in previs, the more useful it becomes to all departments. The important thing for designers and producers to realize is that a centralized previs department delivers a planning tool that can more than pay for itself in both the art and vfx departments.

Which is why a previs unit was started and split up one-third for the art department and two-thirds for the vfx department. We started a previs unit in the art department alongside the concept artist and set designer team very early in pre-production, McDowell continued. But, because there was not a previs unit in the U.K. set up to work within the art department, Warner Bros. and the producers agreed to have [senior previs artist] Ron Frankel come to the U.K. to help set up the team. The previs system was ultimately set up in much the same way as on my other films, but the set up was quite unique for the U.K. Rons prior experience was invaluable, as was that of my U.S. art director, Sean Haworth, who helped make the art department digital and arranged the previs linkages and network. [Previs supervisor] Nic Hatch and his team worked with us to make previs a part of the design process, the storyboarding and the CG planning.

The design sensibility is something that I have to have in place in the early concept period, and although previs is a huge aid in visualizing the logistic issues, its less of a pure design tool as it is a way to visualize the progress of design. I was able to visualize and lay out the design for the two large CGI sets the Boat Ride and the Glass Elevator Ride almost entirely in 3D. The process here was to create a number of concept illustrations for both sets. Dermot Power made some beautiful atmospheric art for the boat racing through the subterranean tunnels of the Chocolate River, and the lighting we worked out in these illustrations became the basis of the lighting of the digital sets. Chris Baker went from design of the Glass Elevator and the mechanics that drive it, to a series of rich conceptuals of vast areas of the interior Factory. We imagined that between each of the Factory Rooms was a negative crawl space through which the mechanics of the Factory, the Oompas and the Glass Elevator traveled. Tim liked the underwater lighting feel of these illustrations, and we tried to avoid defining the edges of the space so that it drops off into a misty atmosphere that has a different feel to the crisp interiors of the rooms.

Although we prevised a number of these early sequences in-house, I ended up working on the final design of both elaborate CG sets largely in the two vfx houses Cinesite for the Glass Elevator and MPC for the Chocolate River Ride. It was a unique insight for me to sit alongside the modelers, colorists and animators working alongside each other and to carve the design directly in the 3D space was very satisfying.

Frankel said his brief role was twofold: I immediately started work on some of the trickier aspects of the film (the mechanics of the glass elevator, the Bucket house) while also training the local animators on [our] approach to previs. It was extremely important that by the time I left London the previs team was working seamlessly with the production, especially Alexs art department. The local animators I recruited had previs and vfx experience and by the time I left London previs had become a central and indispensable resource for the art department and visual effects.

Hatch said previs on Charlie became a central digital hub for the entire production. It was unique in as far as the shear diversity of work we undertook, the way in which we tackled various problems and some of the proprietary software tools we wrote to ease and accelerate the problem solving process. A high percentage of the sets were put through previs for varying reasons. These included the Chocolate River Room, Nut Room, Inventing Room, Marionette Theater, Square Candy Room, Nut Factory, Chocolate Palace, Factory Lobby and exterior of the factory, including the entire town.

Initially we used previs as a visualization tool. Having a physical scale model of the set is fantastic and offers itself well to various discussions. What it doesnt tell you, however, is what its like to experience it from a director of photographys point of view, so we started to animate some of the storyboards to see how close we could get using accurate lenses in the digital set, within the computer. There were huge complications due to the undulating nature of the landscape and the hundreds of candy trees and plants, along with the shear size of the set and the constraints of the stage. We studied the feasibility of using various cranes and camera systems and where exactly on the stage we could place the platforms on which to mount them.

With some sets on Charlie, they would never see the light of day, as by their very definition, they are computer-generated. After talking with Alex, it seemed to make sense for us to lead the design of these sets under his supervision. The White Tunnel ride represents one such set, and is a fine example of how previs helps to tell the story.

Once youve traveled through the Chocolate River Room, you enter into the White Tunnel. A ride taking you down chocolate rapids, past many of the Factorys secret rooms and finally delivering you at the door to the Inventing Room. From the first initial talks with Tim, Alex and visual effects supervisor Nick Davis, it made sense to go straight into prevising the sequence. Tim described his vision, we talked through the script and discussed the key moments, and, based on a few concept illustrations, lead previs artist Martin Chamney [who has formed a new company with Hatch called Nvizage] started to block out the sequence. With no blueprints of the set and no storyboards, its an incredibly creative process. We would show Tim our progress, and very soon we had a fully fleshed out previs sequence that was ready for visual effects shot breakdown and the next stage of more accurate modeling for the final look. It was important that the design language remained true to the initial concept work, so we spent some time with Alex moving that forward before entering post-production.

The same was also true for much of the Glass Elevator ride through the factory. Most of the sets were all CG, but still had to be designed and the sequence conceived, in order to later hand off to the post-production house involved. Again, for the look and language wed work from illustrations and guidance with Alex. We were also hugely concerned that each shot we prevised of the actors inside the elevator were able to be practically shot. Some shots were motion control, whilst others could be shot using more freeform methods. Either way, a lot of thought went into the feasibility of each and every shot.

The Wonka Bar lent itself well to computer-aided design. From initial illustration, we quickly modeled the bar in 3D to a very accurate specification, with a precise Wonka logo on each piece of chocolate. Once approved by Alex, we were able to output to rapid prototyping via an STL file. Three days later, the exact same bar, this time for real, arrived back and was presented to Tim who later approved it.

And how did Burton take to such intense previs? According to McDowell, he started the film with a tendency toward minimal to no storyboarding, and without the desire to get involved in previs. However, when he began to see the previs process that Hatch developed to bring accurate choreography synced to Danny Elfmans music into the 3D set environments, Burton really locked into previs, and began to stop storyboarding altogether. Tim seems entirely comfortable with visualizing actor blocking and camera moves, and would probably not rely on previs if it were only used for its storyboarding capabilities. I think he appreciated the sophisticated ability of previs to gather together all the extremely complex elements into a single production environment.

Not surprisingly, the major focus of previs involved the Oompa-Loompas with all the unique set of problems they introduced. The early problems, however, dealt with planning the Town that surrounds the Factory, and setting up the relationship between the Bucket House and the Factory. The full-scale town, which was built in complex interlocking geographical sections, needed to connect to the full layout of the Town, which would only exist as a model. Within the practical town, built on the Pinewood back lot, were both the full size lower (50' tall x 200' wide) section of the Factory facade, and a large miniature of the factory at the end of a forced perspective street. McDowell said this Factory backdrop had to be lined up with the hole in the roof of the Bucket House roof, through which Charlie sees it, and had to be built in the correct scale and distance to match the miniature landscape. Previs was immensely useful in working this out, and, of course, the data continued to inform the miniature build later in the production.

Probably the biggest single logistic issue in Charlie was that of scale. Oompa scale referred to oompa-sized elements in relation to the animatronic Oompa Loompas, which also applied to 3D props in the hands of CG Oompas; Deep scale applied to settings and props scaled to [actor] Deep Roy himself, for in-camera shots of the actor, and whenever Deep Roy was composited in reduced scale. We also had to plan for several different Oompa conditions Deep in over scale in-camera set, Deep in bluescreen set with over scale elements, animatronic Oompas in human scale sets and CG Oompas whenever they were small enough in frame. If we hadnt used previs from the beginning when discussing the Oompas, it would have been hard for the various departments to properly plan for all the scale issues, and the interlock between choreography, set and camera would have been impossible.

For Hatch, prevising the creation of the four Oompa-Loompa songs was one of the most difficult tasks. The process entailed a number of challenges, notably the way in which they could interactively place 200 or more Oompa-Loompas in an already complex environment while keeping the scene highly interactive for realtime camera moves. If theres one thing thats important to us, its the turnaround time of a shot or sequence and the efficiency in which we go about creating previs. At the time, we really had no idea how many Oompas would be involved in each of the songs, but had to make the assumption that it could easily amount to 200 or 300.

The process started off with Dannys immaculate compositions, one song at a time. Once Tim approved the music, it would move along to choreographer Francesca Jaynes, who rehearsed her routines with groups of professional dancers. On Tims approval we would then video those dancers, rotoscope our previs Oompa-Loompa over the top, duplicate using our proprietary software within our digital 3D sets, and get to work.

We started off with the Gloop song by placing the DV footage of the dancers along with the music into our editing software. From this, we had a timeline from which to work from. We strategically placed some Oompas in the Chocolate River Room and basically starting shooting. Nick Davis and I worked very closely, not least because every single shot in these songs involves visual effects of some kind or another. Tim gave us an incredible amount of creative freedom with everything, but especially when it came to these songs.

Where the complexity comes in, which certainly wont be seen from watching the film, has to do with the size of the Oompas. Each Oompa is two-and-a-half feet tall and every single one based on actor Deep Roy, who is, in fact, four feet and four inches in height. He therefore needs to be miniaturized. So in some way youve got to work out the solution to shooting Deep multiplied by up to 100 or so, all dancing to the same tune, all singing to the same song around the other actors on physical sets, for about two minutes per number.

Nick Davis calculated a number of ways in which to get to where he needed to. In some shots, Deep was dancing on the real set, in others he was on over-sized set pieces, whilst in some he was all CG. But the important thing for us, was that we understood Nicks thinking process and were able to prove his ideas would work before the final decision to shoot were made.

Davis emphasized that they would choose camera angles and moves, with edit lengths based on the natural cadence and flow of the song. The beat and choreography would lead them and as a team they would follow. Slowly they would build up the shape and flow of the song until it eventually became a locked bible for the song, and from this all of the methodology could be devised. Which unit would shoot what; what set pieces needed to be built, as over-scale or bluescreen. This process was repeated four times often with the shooting crew only a week or so behind them. Without previs, I have no idea how we could have even attempted to plan and shoot such a highly complex and technical sequence, Davis admitted. For Tim, it was an invaluable process. He could be completely involved with every creative decision while enabling us to pin him down to exact edits on dance and song numbers before his editor was barely on board. It was interactive, accurate and a bible to what otherwise had been likened to trying to land the Mars rover, while blindfolded at mission control. With no locked, accurate creative plan, the Oompa-Loompa songs would never have existed.

Nor would the Oompa-Loompa songs have existed without the creation of special software. Thus, RODERS (Rapid Oompa Deployment Efficiency System) was conceived. Despite its purposefully ludicrous name, Hatch suggests, this proprietary tool was coded to solve several issues. Essentially, it was made to speed up the process of duplicating mass numbers of animated Oompa characters and to assist their layout in any of our digital sets. We knew that we would have memory limits within Maya and we had to be able to playblast potentially heavy scenes in hardware. Our 3D Oompas were designed with a minimal number of vertices making them light in data, and only a hero Oompa was fully rigged. Once the hero Oompa was animated to copy the dance routines, RODERS would enable us to generate any number of copies, specify path animation, timing offsets, directional options and ID numbering. The transformations of each limb were connected to the corresponding limbs of the animated hero Oompa. The copies would then be positioned on set instantly, and animation driven from the hero. If the choreography in any way, it meant we could easily update the animation on our hero, which would then automatically propagate down to the duplicates.

Other than RODERS, which is actually quite a large amount of code, we wrote a fantastic camera plug-in. Unlike a standard Maya camera, it mimics a real-world moviecam in as far as its basic controls, including various options for Shake. This is far more intuitive for artists to use, but also much easier for a director/supervisor/director of photography to understand when theyre working and sitting with you. The camera also has a HUD (Heads Up Display) showing various user-definable information not just about the camera, but also the entire scene. The idea behind this being, we can create a movie directly from inside of Maya with burn-in, as opposed to having to go through a 2D compositing or editing package. This burn-in tells us and the rest of the crew, scene name and camera name, lens info, frame counter and shot length. It will also display camera height, tilt, pan, roll and speed whilst also controlling the masking and size of the film gate. The HUD grabs all this info directly form the scene, thereby minimizing any human input errors. This burn-in also allows us to track a specific Maya scene within our database for future use and allows on-set crew, via playback, immediate information without having to contact us directly.

Hatch added that they used every part of Maya on Charlie: modeling, subdivision surfacing, animation, shape blending, constraints, trax, rigging, particles, sprites, soft bodies, lattices, hypershade, texturing, expressions and mel.

Meanwhile, the Chocolate River Room was put together using an amalgam of techniques. Without a single straight line in the design, the landscape was designed in three dimensions by sculpting it in scale, in clay. The clay model was then digitally scanned and rebuilt in the previs environment. By this time, there were various storyboarded sequences and a design for the boat with full complement of Oompa-Loompas, so these were added to the 3D landscape. We then tested the camera access to an impossibly inaccessible set, and determined a track for the pink candy boat, MacDowell explained. These elements were incorporated in the digital set and Tim was able to explore narrative and camera in virtual set that included all the action elements. This use was extended when we used the same set to plan the choreography of the dance sequences.

In the Chocolate River Room, previs was used to help decide the number and placement of the Oompas in the boat, the path of the boat, the Oompa choreography that evolved with composer and choreographer, and the 3D Oompas were placed specifically in the set so that all camera moves could be planned. Over-scale Oompa synchronized swimming sequences were prevised to plan the necessary sfx rigs. Set extensions were designed in 3D and color coded in the digital landscape for vfx planning, and the animation and scale of the sucking machine that sucks up Gloop was resolved in previs before going to CG modeling. Working out the landscape design of a large variety of new flora before they were fully built was a concern both to Tim and I. We actually constructed a catalogue of plants in 3D that we could place and move around the digital landscape to determine not only the relationship between the plants and the landscape, but also how the camera would negotiate the taller trees. Similarly, in the Nut Room, the number of squirrels, the planning of 3D vs. animatronic vs. real squirrels, the decision to eliminate set extensions, etc., was all decided through the use of previs.

The Glass Elevator, of course, provided its own unique challenge. In fact, the idea for the Elevator was not to be a magical object but an unusual kind of gondola that respected most of the laws of gravity and engineering. Very early on, according to McDowell, Frankel set up a test elevator and track to work out how the Elevator would move through the Factory space, and how it would react to abrupt changes of direction when it flies sideways as well as up and down. This formed the basis of the elevator animation, which Hatch and his team continued to refine, although they found that the simplest rail mechanism worked the best, in keeping with Burtons general approach.

The unique challenge of the Elevator was its complete transparency, McDowell said. Much of the time it was shot against bluescreen and the reflection issues were interesting, but in wide shots the elevator was generally a CG element, often with CG characters inside. With the previs sequences, Tim, [cinematographer] Philippe Rousselot and Nick were able to plan the number of shots and the types of camera moves. As the shooting progressed, Tim reassessed and changed the number and size of some of the Elevator Ride sets, and we had to do some rapid redesign of in-camera elements as well as miniature and CG sets. The Puppet Hospital, the Exploding Candy Room, and the Candyfloss Sheep were all sets that were added after I had finished most of my designers, so they were designed directly in previs. In a sequence as abstract as the Elevator Ride, establishing a geographical logic is important to the audiences sense of space. In previs, and at Cinesite, we created a 3D map that gave everyone a sense of direction, and helped our AD, Katterli Frauenfelder, impose the same logic for the in-camera shooting.

Hatch opined that previs has become an overall production tool, not just in pre-production and production, but what they now refer to as postvis. And Charlie was no exception. There were a few sequences that had been shot, but lacked the vfx needed to aid in the telling of the story. Wed very quickly gather the dailies from editorial, rough-track the plate, if necessary, and insert the visual effect, be it Gloop in the river being sucked up the pipe, or Violet turning into a blueberry, Hatch explained. Where this really helps the editor and director, and ultimately everyone else involved, is in the pinning down of what the shot and sequences require to make them work as a piece of narrative. Often post-production houses are not geared up or able to complete the postvis work quickly and efficiently as they are more intent on, and quite rightly so, finishing work to a finite polish, which will always take longer and require a different skill set.

The challenges also involved working with special effects (physical effects) supervisor Joss Williams, where we helped determine a number of issues, including the best path for the boat to travel down the river, both mechanically and aesthetically, without either the boat or its oars hitting the side. It was also decided that the boat would need a turning point. We used previs as an efficient tool to determine where best to re-model the landscape to insert that turning point. Previs artist Jon Graham sat down with senior art director Kevin Phipps, calculating and re-sculpting the organic grassy landscape within the 3D software. Drawings in the form of topological lines were given at intervals of two feet to construction for them to start the huge task of the physical build. Previs artist Jarrod Linton also helped work out the most efficient way to capture Gloop stuck in the pipe, whilst also calculating the necessary size of the bluescreens and their exact placement within the stage.

Whilst demonstrating to numerous members of the crew via the computer monitors how wed solved various issues, there is also a definite need to create printable diagrams and layouts showing precise camera placement and lens information along with the most efficient use of stage space in certain sets. Previs lends itself well to working out false perspectives and painted backings along with green and bluescreen size and placement. These diagrams become excellent discussion points in meetings, whilst out on set or on location.

In essence, we ended up with a previs department answering to the needs of numerous crewmembers. In fact, on a film we were recently part of, called Watchmen [that has since been shut down], thats exactly what happened. It made total sense to have a centralized previs department, which answered the needs of the director, Paul Greengrass and all departments. Paul saw it as the films digital nexus, which made complete sense. With the correct team and tight scheduling, that setup works very well, and is something I see becoming more embraced on the projects where it makes sense for that to happen.

What all this means is that previs is now increasingly becoming a central part of the entire production, with more and more directors using it as another tool to envisage their creations.

Bill Desowitz is editor of VFXWorld.