Pandora's Box

 

 HOW Magazine, August 1993

 
 
 
 
Computers have changed the way we live, design, produce and think. When they were expensive and difficult to operate, their impact was limited: but the last few years have been a watershed. Computers' prices have dropped , while their power and capabilities have soured. The proliferation of these "calculating" machines makes it impossible to conceive of  a world not shaped by the computer. But the old question remains:  

Where do computers fit in the fabric of our lives? 

 
 
 
 

  
 
 
As one of the world's best known special-effects creators, R/Greenberg Assoc. (New York City and Los Angeles) confronts this issue every time it embarks upon a new project. Whether resurrecting Gene Kelly to dance alongside Paula Abdul in a Diet Coke TV spot, or zapping a young boy from the theater audience smack into the middle of a superhero action movie in Arnold Schwarzenegger's "Last Action Hero," the company's livelihood is contingent on its knack for using the computer to make the unbelievable seem real. This is done both through its film division and through R/GA Print, the firm's digital print division established to apply that skill to still imagery. 
  
R/GA Print brought that expertise to this issue's cover. Armed only with the physical specifications and the stipulation that the cover had to convey the computer graphics theme, the creative team was free to address the headier issues associated with that topic. 
  
Joe Francis, a director at R/Greenberg Assoc., came up with the concept. "I didn't want to create just another pretty picture. I wanted to design something that had meaning beyond the picture," he says. 
  
His idea, inspired by the Greek myth of Pandora's box, was a play on our relationship with technology. The cover also addresses questions inherent in technology by demonstrating that its use is our responsibility: Technology is neither good nor bad; rather, the way we employ it is the issue. Technology should be used to communicate the concept, not be the concept itself. 
  
Francis and his collaborators chose to let the creative process itself point us in the right direction. From the beginning, they made a conscious attempt to de-emphasize the sharp, hyper-real look that characterizes much computer generated art. Instead, the image's fulcrum is the contrast between 2D and 3D realities. Pandora appears in two dimensions--flat on the side of the box, and 3D when her hand reaches out to remove the lid. She plays two roles in the image--she opens the box, but remains a part of the box itself. 
  
  

IN THE BEGINNING

Francis collaborated with photographer Dan Wilby and Bob Bowen, manager of R/GA Print, to develop the idea. He began by making a 3D stage on the computer using software he wrote specifically for the project (fig. 1). This stage later served as the framework upon which the photographs were rendered. The computer grid determined how the set for the photo session was built and the model positioned. 
  
Next came the photo shoot itself. Francis painstakingly explained his concept during the session, so Wilby could effectively capture the mood. The photograph had to represent the Greek myth of Pandora as she opened the box entrusted to her, as well as show the natural curiosity associated with scientific discovery. Wilby otherwise had the freedom to make aesthetic decisions about how the photograph should look. By carefully controlling the light and using filters to soften the edges, he satisfied Francis' requirements. 
  
Wilby first designed the set with a prop artist, then set up the scene with lights, using his assistant, Barbara Norman, as a stand-in for positioning. Everything was in place before the model arrived. After an extensive makeup session (done, ironically, to achieve a natural look on film) and a Polaroid test, Wilby shot several rolls (fig. 2). The scene was then recreated without the model to "build" the sides of a box. 
  
The photographic integrity of the finished picture required careful lighting. Illumination makes a photographic image appear real and is often the bane of 3D imaging. Digital rendering cheats in the way it constructs light: 
  
Light reflects backwards from the digital camera to the source, instead of the other way around. This reverse engineering requires various techniques to make the light appear realistic. 
  
Wilby placed a light inside the box on a table (fig. 3), and its shadows became a crucial element in flattening out the 3D look of the model. The shadows were later digitally multiplied into the model's photo to make it appear as if she were a 2D picture on the wall of the box. 
  
The photographs were then scanned at 4K resolution. Small test copies were applied to Francis' 3D model and digitally rendered. 
  
The first tests showed that multiplying the shadows into the image did not adequately transpose the model's 3D look to two dimensions. So the creative team tried to create a more convincing 2D effect by adding texture to the wall. The biggest problem with computer-generated images is making them look real--but not too real or sterile. In this case, however, there was an attempt to introduce organic aberrations and other optical distortions into the image to maintain an analog look. So they tested various surfaces, scratches and cracks with the idea that the aberrations would fool viewers into thinking the model was a 2D image. But that idea was ditched when the aberrations proved distracting. 
  
The second attempt was more successful. Francis created molding that runs over the model to flatten her out. He then increased the contrast of the photo of the lighted box to bring out the 2D shadows. Combining the resulting image with that of the model caused her image to darken and I warm considerably I (fig. 4). 
  
As the preproduction stage was coming to an end, the image determined its own fate. Rather than cropping in on the I action, Francis kept zooming out, drawn to the warm tones of the model's skin in the background. 
  
The picture seduced Francis to further develop his initial concept, which changed from a mere allusion to the infinite chain of boxes within one another, to a literal representation. Wilby was present during this transformation and was amazed to see his photography evolve before his eyes. "There is no way that could be done photographically," he says. 
  
R/Greenberg's 16 parallel processor IBM POWER Visualization System supercomputer crunched through the production and delivered a 4K file in less than an hour. The rendered file was a rough composite of the finished piece (fig. 5). Francis eventually refined and re-rendered sections of the image, but the remaining work would be done in postproduction. 
  
  

FINE TUNING

Post-production took place on the Macintosh with Adobe PhotoShop. Bowen, Karen Sideman and Frank Lantz integrated the rendered image into a finished picture (fig. 6). Though R/GA Print has Mac Quadras with hefty 2GB RAID drives and 128MB of RAM, working on a whole file at once is a painfully slow experience.  There might be up to 20 versions of the same file, which can easily absorb nearly a gigabyte of disk memory, so ample disk real estate was a necessity. Whenever possible, the area to be worked on for the HOWcover image was isolated into a smaller file. 
  
A lot more work had to be done beyond compositing the rendered images. The 3D geometry was cleaned up, subtle lighting changes added and the image unified. 
  
An image of this complexity has to be handled in a logical order. "It's similar to chess," says Bowen. He began by positioning key elements in the frame. Once their position was determined, he organized the image and divided it into tasks that Lantz and Sideman could work on in tandem. These parts were composited back into the image, then saved at each stage so it was possible to go back, in case something needed reworking. 
  
Bowen first touched up the background. He used various masks and the Levels command. Bowen, Sideman and Lantz then slowly moved to the foreground, retouching each element as they went. According to Bowen, "Working from back to front and from front to back is the secret to remaining flexible, for keeping the transparency of effects, and allows for graceful repairing of problems as they come up." 
  
Francis then re-rendered problem areas that couldn't easily be corrected in PhotoShop and spliced them back into the original. 
  
Photoshop's selection tools, which provide precise control over the work area, enabled Lantz to easily light the image. To add a soft light to the left panel in the closeup (fig. 7), for instance, he first created a selection using the Lasso tool with a feather radius of 15-20. This kept the transition smooth when he either increased the brightness or raised the gamma. He then used the pen tool to eliminate any selection area extending beyond the box. 
  
Further refinements at this point included changing the direction of the inner box and adding another box. 
  
In readying the picture for print, Bowen inspected the image at a one-to-one resolution, making sure there were no aberrations. "Any imperfections that are visible are guaranteed to be there in the final output," he notes. One way to inspect the image is to turn up the gamma using the control panel device Adobe ships with Photoshop. This makes it easier to see detail in areas where flaws might otherwise be hidden. 
  
  

FINAL OUTPUT

The picture was imaged to an 8xlO-in. chrome, then converted again digitally for separation. This digital-to-analog-to-digital process, as opposed to simply handing the file to an output house, has several advantages, says Bowen: The algorithms for translating film to print are optimized and predictable; color remains consistent with the artwork; art directors can evaluate color on chromes; and clients can reuse the picture, change its size or crop it as needed. Calibrating monitors is also easier because the image remains in RGB color space throughout its digital life. 
  
The final image is a representation of reality--with a twist that catches the viewer on a subconscious level. The illusion of two- and three dimensional space creates an impossible reality, a fantasy that illustrates the relationship between humans and technology.  
  

  
 
sidebar: 

ABOUT R/GREENBERG

R/Greenberg ASSOG. is a bicoastal company famous for its award-winning special film effects, seen in the Diet Coke ad campaign in which movie stars of the past are brought to life alongside today's hottest stars, and Arnold Schwarzenegger's "Last Action Hero." 
  
In 1989, company president Robert Greenberg formed R/GA Print, a digital print division. Under the leadership of Jimm Burris, the division has created the capacity to apply special effects developed for film to print.  
  
The integration of a computer-generated 3D world into believable pictures gives images a memorable dimension. It is also the current trend in special effects, "which is no, or transparent, special effects," says Burris. An effect should be absolutely seamless so you believe it--and only after you believe it do you stop and think, "But that's impossible." To see a Kawasaki dirt trike pull a lassoed tornado looks completely ordinary until you realize it's not possible. Burris says, "The hardest part is explaining to a client what we can do, because we can do anything." 
  
The re-creation of reality requires teamwork, from concept development to photography, 3D rendering and integration of the elements in postproduction. R/GA Print brings together photographers, designers, illustrators, model makers, soft-ware engineers and computer artists all under one roof. 
  
Most 3D work is done with R/GA's own proprietary software written in C, which allows R/GA Print's computer effects artists to custom program their solutions. Though they do use some off the-shelf software, such as SoftImage, they usually write code to produce a specific effect. 
  
The digital-imaging programs were originally written to create effects for film, but they are just as applicable to print. 
  
.