In Moviemaking, Computer Graphics is Turning Modelmaking into a Lost Art
The problem with a new technology is that, sometimes, old ones
must die to make room. Such was the case in the making of the
Tim Burton film, Mars Attacks! In earlier films like A
Nightmare Before Christmas and James and the Giant Peach,
Burton had used a time-honored animation technique called stop-
motion, in which inanimate objects are successively moved--
slowly and carefully--one frame at a time. When shown on a
screen at 24 frames per second, these successive frames can
turn a figure made out of plastic, wood and rubber--a figure
crafted by a modelmaker--into a lifelike creature.
*One of the best sites of Tim Burton is here.
And so when it came time for the director to produce green aliens for his science fiction spoof, Mars Attacks!, Burton turned to Barry Purvis, an English stop-motion filmmaker, to head an international team of animators. The team's early results were certainly successful--Burton liked what he saw. But the time- consuming stop-motion process threatened to break the film's budget, which the studio wanted kept at $70 million. While the stop-motion team worked away, Burton hired George Lucas's now legendary special effects firm, Industrial Light & Magic (ILM), to produce its own version of the aliens with its stable of Silicon Graphics workstations. Jim Mitchell, who supervised the effects for Jumanji, produced a five second computer-generated test showing Martians pulling a bumper off a car. The results were good news for the Mars Attacks! budget and bad news for Purvis and his crew, who found themselves unemployed.
It was one more triumph for the science for computer graphics (CG), whose photorealistic images now fool the eye into thinking they are the stuff of form and substance. These techniques include high-end rendering, ray tracing, the computation of shadows, and shaders used to make surface bumps and divots. These effects are produced by a handful of software packages, most notably, Alias|Wavefront.
Another example of CG's move on model making: the updates to the Star Wars trilogy, including a new footage of a younger Jabba the Hutt, and Dragonheart, a film whose lead character, a robust looking dragon named Draco with the voice of Sean Connery, is completely computer generated. In the new Star Wars trilogy, which is due in theaters beginning in 1999, about half of the scenes will be created through computer graphics at an estimated cost of $60 million, approximately half of what it would have cost using traditional modeling techniques.
Models and computer graphics: mixing them together
In filmmaking, computer graphics doesn't always dominate;
sometimes, it shares equal billing. During the production of
Jurassic Park, for example, a young special effects wizard at
ILM, experimenting on his own time, created a herd of
galliminus as well as a walking tyrannosaurus rex. The demo
convinced director Steve Spielberg that CG could indeed stand
in for models, at least in some shots. A good rule of thumb
for Jurassic Park is that the closed ups and medium shots are
models, while the long shots--particularly those of herding
animals--are CG. That mix has continued on the Jurassic Park
sequel, The Lost World. "Lots of people suggested that The
Lost World should be all CG, because now they've done all-CG
movies like Dragonheart and Toy Story," said Michael Lantieri,
The Lost World special effects supervisor, to the U.
Another film that uses both digital images and models is Independence Day, which relied heavily on large-scale models of such American icons as The White House and the Washington Monument--assembled in several former airplane hangars in an industrial park near the Los Angeles airport. Dean Devlin, producer and co-writer of Independence Day, told Premiere: "Digital technology, especially for computer-generated animation, is a fantastic tool, but it's not for everything. There is a tendency to latch onto whatever new toy is invented, but sometimes you need a good old-fashioned model on a string."
A matter of money
But while modelmaking is not dead, budgetary considerations have severely reduced its use. Back when Spielberg made Jaws, for example, computer imagery was never an option. The shark was played by a succession of mechanical models (collectively named "Bruce" after Spielberg's lawyer, Bruce Ramer), each weighed3000 pounds and cost$150,000 to build. Fast-forward 20 or so years, and the costs have only gone up. Spielberg's latest film, The Lost World, employed two T-Rex models--his and hers--that weighed about 15,000 pounds, were controlled by up to 10 puppeteers, and cost about $1 million each to build. Not surprisingly, if filmmakers can find a way to create the same effects in software and silicon, they will do so.
Some of the best known computer work is being done at two locations: Lucas's ILM, which sits in a series of unmarked buildings in the industrial area of San Rafael, California, about 20 minutes north of San Francisco, and at Pixar, which is in the city of Richmond, just across the San Francisco Bay. In some ways the companies are siblings. Pixar began as a Lucas-owned company and was sold to Steve Jobs in 1986, with Lucas getting all software advances made by Pixar royalty free. (Lucas did a similar deal in spinning off its digital editing system to Avid Technology.) Like Steve Job's other company, NeXT, Pixar first tried to sell it's own workstation into the graphics niche now dominated by Silicon Graphics-- before eliminating the hardware and concentrating on its software.
But Pixar had something else up its sleeve: Academy award caliber filmmaking talent talent. A small group of technicians within Pixar, headed by Disney-trained animator John Lassiter, were producing one-of-kind computer-generated films that were initially intended to show off the prowess of the Pixar workstation. The first of these was Luxo, Jr., which featured two desk lamps and a beach ball. Running less than five minutes, Luxo, Jr., with its interplay of light and shadow, remains elegant in its simplicity and a dramatic testimony to the power and potential of computer graphics as an entertainment medium.
In time, Pixar's small animation group became the force that now drives the company. Its first full-length motion picture-- Toy Story--is a virtuoso performance for Lassiter and his team. Toy Story has about 75 characters, 360 objects, and 1700 shots. The film's 109,000 frames were created with a combination of Alias|Wavefront software and Pixar's MenV procedural modeling environment on SGI workstations, and rendered on a "farm" of 300 Sun machines, which computed the color, lighting and texture. The results are so realistic that many film goers may not have realized that none of these simple toys "exist" in the usual sense of the word.
Computer graphics grows up
The technology to do this has been a long time coming. "Photorealism has both a scientific and an artistic component," said Peter Shirley, assistant professor of computer sciences, University of Utah, which for many years has been considered a hotbed of research into computer graphics. "A good artist with the right tools can make something look pretty realistic just by painting it with light in 3D. Pixar pays a lot of people a lot of money to do that for a lot of hours at a time."
Photorealism depends largely on the ability of software to cross a few big hurdles--partly in producing more accurate image geometry, and in part by generating more realistic surfaces. In some cases, a clay model can be scanned in using a 3D digitizer, using a laser or manual probe to define 3D coordinate data.
Lighting is another challenge, particularly indirect lighting in which the effects of a say, a halogen light bouncing off a ceiling, is replicated. "Indirect lighting can make you feel that an object is real," said Shirley, "but right now that's not supported in most commercial packages. If you want the effect, you must create it by hand. That's why a lot of computer graphics looks more like an illustration than the real thing. It still takes an artist to make something truly look photographic."
Getting the lighting right can be surprisingly difficult for filmmakers, too. The flying saucers in Mars Attacks!, for example, were modeled in only a few days. But even though they are symmetrical, smooth images, they proved a challenge, requiring a complex layering of shaders and textures to give them the look the filmmakers sought--reflective, but not too reflective--with a suggestion that they are covered with a thin layer of dust. Similarly, when the saucers land in the desert, the image makers had to create different environment masks to accommodate the changing weather, which alternated from sun to rain on the days they were shot.
Hence, while software can accurately replicate a chrome bumper reflecting the woods behind it, it is much more difficult to model the lighting of the interior of that same car. And if you add a driver to the scene, you lose the photorealistic quality entirely. Shirley calls the creation of photorealistic images of people "the grand challenge problem" of computer graphics. "Not only do people have very complicated geometry and material properties, but real humans are great at spotting mistakes. If I show you a computer- generated dinosaur or toy robot, you won't notice a few imperfections. But the same mistakes in a person would be immediately apparent. I don't think actors are going to be out of business in our lifetime."
But others are not so sure. Some observers are pondering the question of copyrights on computer generated imagery. What happens, for example, if you create a life-like model of Marilyn Monroe? Can you co-star her with Tom Hanks? Does Monroe's estate get to collect? The temptation to resurrect the dead, and to replace Hollywood stuntmen with more expendable computer-generated stand-ins, will keep the artists who wield computers firmly fixed on this, the toughest CG problem of them all.
An interview with Clint Goldman, Producer, Spawn
A producer and executive producer at George Lucas's Industrial
Light & Magic for seven years, Clint Goldman has taken a leave
of absence to produce Spawn, which is due out in the U.
I spoke with Goldman by phone from his office in the Todd-AO building in Hollywood, where he was "in post-production big time."
- To what extent are computer graphicsovertaking model making?
- In terms of its prominence in the motion picture business, model making is not what it used to be. ILM's model shop, for example, is still quite extensive, but it's not their bread and butter any more. But then you get a movie like Independence Day, which was made primarily with miniature pyrotechnics using large scale miniatures that are digitally composited. ID4 [Independence Day] stole the year last year, even though Twister was far more computer generated.
- The truth is there is a combination of those techniques that are still as successful as others. However, most of the effects in Hollywood are being done digitally, and it will continue in that direction.
- Is there a rule of thumb deciding where and when computer graphics will be used?
- Well, stop motion is gone for the most part, with the exception of a fully animated stop motion film like what Henry Selick does on Nightmare Before Christmas or James and the Giant Peach. Stop motion was responsible the skywalkers in Empire Strikes Back, and a lot of the work that Phil Tippet did at ILM on RoboCop--but its an art form that has just about disappeared. You just don't see that style of work any more because of the computer. But for a space ship in Star Trek or Independence Day--you can detail a model quite well, and then move it through space. The key is that the ship doesn't have to animate within its own element.
- Will models only be used when they are blown up?
- In terms of explosions and miniatures, you can still do the work just as well and probably cheaper and faster using some sort of combination of the two techniques. Even though people in the industry don't think that the work was done very well, Independence Day was a big, slick, fun summer spectacular-- with multiple airplanes and ships that were bigger than Manhattan. Those were miniatures, and to a lot of people, they made for a good spectacle.
- But as the generations change, there won't be as many skilled modelers. Ultimately, I think all the work will be done digitally after this generation of modelmakers has left the business.
- A lot of people mark the point of transition between models and computer images with the creation of a digital T-rex on Jurassic Park.
- That's true for the public, but in the industry, the first glimmerings came with the stained glass man in Young Sherlock Holmes. There was a lot of work like that that was pretty good but not significantly noticed. The pseudopod in The Abyss, for example, had a strong impact on the industry.
- I think that the filmmaker who had the biggest impact is James Cameron. Terminator 2 (T2) was the seminal film because it made so much money and showed the power of computer animation. Jurassic Park was the film that made it clear to everybody in the world that there was no limit any more to what could be done with a film. But T2 was definitely the film that got the industry to believe that it *could* be done.
- But not everyone believed it.
- That's right. Even in the beginning of Jurassic Park there were naysayers at the very highest level of Industrial Light and Magic that believed creatures couldn't be created digitally. But at that point, there was enough steam internally at ILM that it was going to happen.
- There are two huge examples of how computer graphics hasencroached on model making. One is Jurassic Park, where Phil Tippett with a large crew was going to do a number of the herd shots, running dinosaur shots as stop-motion animation. As soon as Spielberg saw the test that ILM did of the galliminuses running--where ILM would take a walk cycle and then offset it, and then duplicate the number of galliminuses in the herd--he was convinced that that section should all be done with computer animation.
- The second example is Tim Burton's Mars Attacks!, where he had assembled a stop motion crew from all over the world. ILM did a test using computer animation, but with a different technique to give it a more "stop motion" look. It was uncharacteristically quirky for computer animation and Tim really liked it. He felt CG added value that he couldn't get with stop motion. So they laid off the modelmakers.
- In that case, did it all come down to the budget?
- The big problem was that the time factor involved in doing the stop motion pushed the release of that movie back a long way-- further than November of 1996, which is when it came out. The interest [as in interest taken out on a loan] costs on these movies are absurd in terms of missing your delivery dates. People are sometimes more interested in spending more money to get a movie out quickly because the actual costs are more if you extend the interest out on a $100 million investment.
- But while money typically drives most of the decisions, the creative element is the still the strongest element in Hollywood. Tim Burton liked that technique more, and when Warner Brothers got the news of what it would cost and how long it would take, it met their demands as well.
- What about in your own work? Down the line are you still going to use both?
- On Spawn, we're trying to do an incredible amount for the money we have to work with--trying to make a big budget effects picture even though it's a smaller picture. But the effects will be big. There was one instance in particular where we chose to do a rod puppet creature, because it was only going to be in five shots in the movie and we knew we could cover it up with digital fire. But as it turned out, when we edited the picture his role became more significant and everybody was getting nervous that it wouldn't cut the mustard.
- If finance hadn't been an issue we would have definitely built the creature digitally. But at the time we didn't have the money to do it. So we ended up expanding this digital hell sequence, which is a significant part of our movie. There are four sequences spread throughout the movie that are all short- -20 seconds or so. At the end of the movie, there's a three minute sequence with this creature. We were able to round up enough dollars to give this particular company the right amount of money to do the sequence properly. They got motivated since they were doing the entire digital hell sequence and they decided to go ahead and model the creature on their own. They built it digitally, and it's going to make all the difference.
- People like myself and my two partners are going to push it into post production whenever we can. For example, the budget for our movie is $40 million. More than 25 percent of that is visual effects, and almost half of it is in post production. My hunch is that our next movie will be more than half post, and perhaps close to 40 percent will be visual effects.
- Besides modelmakers, the other people who have been replaced in Hollywood are stuntmen. Not completely replaced, but we definitely made choices while we were shooting to not spend the time or take the risk with doing something with a stunt, knowing that it could potentially cost us more money to do it later on digitally, but also knowing that if we did it digitally it would be safer, and ultimately we would get a better result.
- Isn't one of the assumptions of computer graphics that it still can't simulate people?
- The truth of the matter is you can do anything now, given time and money. We have four digital characters in our movie. They are not human beings yet, but we have this actor playing Spawn, who wears a rubber suit. and we've matched the look of the rubber suit identically, and he's got all kinds of musculature on him.
- We know how to cheat things. If you are a smaller production team like we are, making a movie for $40 million instead of $100 million, you can't push things that aren't ready to be pushed yet. Even though you might be able to make it work ultimately, it will cost so much that it will hurt the movie overall--you will have spent too many resources in trying to make something work.
- For example?
- Say you are on a wide shot of a guy falling off a building. Our character is suited with armor that has a mind of its own, with chains that come out of his chest that he flips over. We could have probably gotten by with a human doing this, but it wouldn't have looked like this particular character. So going digital not only gave us a better performance, but produced something that was more in keeping with the character's persona.
- The technique also ensured that we could tweak the performance. We weren't worried in a tight shot or a wide shot that we couldn't get again. When you're on a set with lots of people and only 12 hours in a day, there's often no room to go back and do it again. By contrast, with post [production], you don't need to make it happen over a day. You can take a month to make a digital character work.
- There's lots of scenes in our movie where we'll go from a practical [actor-played] Violator, who is our beast, and a practical Spawn, cutting right back to a digital Spawn and a digital Violator. If you really wanted to, you could tell the difference, but the effects are good. Most of the time the digital stuff is better than the practical stuff--it looks better and has more flexibility.
- It sounds, in some way, like computer graphics is taking center stage in films. Is that a good thing?
- Every single big movie that came out last summer had a lot of visual effects in it--every one. And the same is true this year: Anaconda, Men in Black, Contact, Titanic, The Lost World, Dante's Peak, Fifth Element, Volcano, Air Force One, Starship Trouper, Alien 4. Non-visual effects movies are no longer the big summer movies.
- We happened to grow up in this era where you can create a whole world of stuff that audiences haven't seen before using post production techniques. Those are the kinds of movies we are making. It's not right for everyone, but it's right for us.