DreamWorks Goes on a 3-D Rampage with <em>MVA</em>

[1]

Monsters vs. Aliens [1], DreamWorks' first 3-D feature, pulled out all the stops to create a fun, broad actioner. All images © DreamWorks Animation LLC.

Now, after all the 3-D hype, Monsters vs. Aliens [2] finally arrives in theaters today from DreamWorks Animation [3] (released by Paramount). But Conrad Vernon (Shrek 2 [4]), who helmed the feature with Rob Letterman (Shark Tale [5]), is the first to emphasize that MVA is not a 3-D gimmick in its depiction of a gaggle of monster misfits out to save the world from a maniacal alien.

"What we talked about from the very beginning, before we knew it was going to be 3-D, was doing something a lot bigger and more spectacular in terms of scale than had ever been done before," Vernon suggests. "There's room in animation to do the beautiful, little, quiet films, like Ratatouille [6], which is character-driven and one of my favorites. But we wanted to blow the roof off… So when we first started thinking about doing this in the old Ed Wood, B-movie style of science fiction and horror, it was fertile ground to satirize these films [Tarantula, Godzilla, Attack of the 50 Foot Woman] and have some fun with them. We also wanted to make the action sequences real action sequences. We wanted there to be real danger, real stakes and spectacular visuals. And DreamWorks let us go with that..."

[7]

The directors aimed to "blow the roof off" with Monsters vs. Aliens [7], giving a palate for the 3-D to work on.

However, this ambition to pull off the biggest actioner in the history of DreamWorks Animation began with a design style inspired by Mad magazine. "I gave Craig Kellman, our character designer, the Jack Davis template and he went to town," Vernon adds. "We didn't want to rubberize the characters too much. We wanted them to be funny but we didn't want them to break bones or have eyeballs popping out. For instance, I gave everyone the idea of John Cleese doing the funny walks from Monty Python. Look at that for reference for how the President moves. He's almost falling down constantly, but very elegantly picking himself back up."

As far as the animation, Vernon wanted to make sure it was unique. "I wasn't dictating what the animation should be, but telling them to break out of the normal little poses and get into the characters' emotions. And I would actually get up in front of the animators and act things out for them. They would ask questions and we finally settled on a style."

According to Dave Burgess (Bee Movie [8], Madagascar [9], The Lion King [10]) head of character animation, "The directors and I kept going back to Chuck Jones [11] and reference how his characters really milk their poses. They'd stay in their poses for a long time and there'd be secondary actions with hands and shoulders and little shifts of the head. But we didn't feel the need to illustrate every single word or phrase being said in the time-honored way of having it done in pantomime.

[12]

Characters were to hold their poses for a beat, giving the animation a Chuck Jones feel that should feel unique. [12]

"Personally, I have an aversion to typical acting you see in animation where you [hold very long poses]," Burgess adds. "I told the animators they would have to figure out something else to do."

Overall, Burgess thinks the CG humans are the best yet at DreamWorks, thanks to the latest technical advances as well as a mandate from the directors. "There's a depth to the skin and a level of detail that isn't photoreal but it blends with the design-y characters. Again, I think we did a good job of avoiding the Uncanny Valley. They're stylized enough, especially Susan/Ginormica (voiced by Reese Witherspoon) and Derek, her fiancé (voiced by Paul Rudd), but also General Monger (voiced by Keefer Sutherland as a cross between R. Lee Ermey from Full Metal Jacket and Yosemite Sam) and the President (voiced by Stephen Colbert) as well, so we were able to find a fairly sophisticated look that augmented the design.

[13]

The Uncanny Valley was deliberately avoided in the human characters, like Susan (pictured with The Missing Link), who was stylized but had a level of detail approaching photoreal. [13]

"We worked very hard in developing the translucency of the skin. With Susan and the President and Monger, we added wrinkle controls to their brows and to their noses. We could move the skin a little before but now more than ever. One of my pet peeves with CG humans is having too much dead space between the nose and eyelids. There's this little rectangle that is hard to move around. A lot of it was directorial choice. We wanted to have more of a living, breathing presence, and with all the surfacing and other stuff, it blended together really well. We had more bendy arms to break the bones between the wrist and elbow. We could actually bend that into a curve, if we wanted for in-betweens."

Meanwhile, the gelatinous B.O.B. (voiced by Seth Rogen) posed the biggest character technical challenge. "We spent at least a year in development on him," Burgess continues. "Initially, it was a mystery whether he would be an effect or a character rig. And we ended up with a hybrid: a basic character rig with blobby effects on top when rendered. When we worked on him in our animation software (EMO), we didn't see the goop at the base, his translucency or the bubbles until we got renders back from the farm, where all that was generated."

[14]

B.O.B. took a year of development, and ended up as a hybrid character rig with blobby effects. Here, he charms the pants off some Jell-O. [14]

The Missing Link (voiced by Will Arnett) and Insectosaurus were unique-looking but not different in terms of rigging. But Burgess had a mandate for the TDs to get a tail system that dragged, more in keeping with the Japanese monster movies they were referencing. "Insecto is 350 feet tall and when we were working in stand-in modes in EMO, we didn't see that fur thickness," Burgess explains. "For instance, there's one place where Susan backs up into him and touches him with her hands. By the time that we got the fur turned on in that shot, you couldn't even see Susan, she was so deep into the fur, so we had to cheat visually for stuff like that on a shot-by-shot basis."

Dr. Cockroach Ph.D. (voiced by Hugh Laurie in his more natural English accent but with a touch of House by way of Dragnet) was an exercise in contrast as the mad scientist. But temperamental alien Gallaxhar (voiced by Rainn Wilson) was more of a challenge once they got into the facial and tentacle work. "Adding two more eyes to the face was a pipeline adjustment," Burgess offers. "The tentacle thing was a rebuild. Instead of having him walk like a big spider, we had him walk like a stalk of string cheese that has a bit of the cheese shredded at the bottom. It was another instance of a Chuck Jones-like feel to the movement. I did a rough 2D test for all of the monster characters early on to help the riggers, and the directors [especially] liked what I did for Gallaxhar."

[15]

Evil alien invader Gallaxhar was animated to look like a piece of string cheese shredded at the bottom, instead of going the spider route. [15]

Of course, it helped getting a huge bump half-way through production with an improved EMO. "They sped up geometry caching so you could turn on a lot more geometry in your camera view and spin it around and see it move in space," Burgess explains. "When doing blocking, there's a way to render it around eight times as fast. We also got webcams to all the animators [for the first time] to better shoot video reference at their desks."

And when it came time to executing the bravura chase in San Francisco and subsequent confrontation on the Golden Gate Bridge, Digital Effects Supervisor Mahesh Ramasubramanian (Bee Movie, Madagascar, Shrek [16]) was well armed. In fact, MVA was so grand in scale that there were two teams supervising lighting while he separately supervised the chase and Golden Gate Bridge sequences.

[17]

The Golden Gate Bridge set piece involved a 50-foot woman, 300-foot alien robot, 350-foot furry insect and a bridge that was the largest piece of rigging DreamWorks had ever done. [17]

"Basically, this was an opportunity to splurge on effects and really have an impact," Ramasubramanian says. "The Golden Gate Bridge involves fight between the monsters and the alien robot. We made this sequence as grand as we could and the directors decided to base this sequence in California and in a very iconic location. We first went about building this whole bridge and it was very detailed. We found that if you reproduce all the detail, it is too heavy, so we went in stages. First we built the model without adding too much detail and kept adding more based on what you see. But what was important here was for us to get a sense of the scale of the different elements of the shots. We had the bridge, which is immense, of course, we had Susan, who is nearly 50-feet tall, the alien robot, which is 300-feet tall, and Insecto, which is 350-feet tall.

"After we built the bridge, we realized that it was more of a character during this whole sequence and not a set piece. The bridge was being moved by the characters and the rest of the crowd in the cars had to react to the bridge. So what we did was rig the whole bridge. We didn't leave it as an environment piece that was later destroyed by effects. We put a rig on the bridge so character animation could animate it like a character and create all of the gross movement. This turned out to be a pretty big task because we've never undertaken rigging as big as this. Fortunately, our tools have advanced enough that we could do it: this allowed everything to be done in animation rather than going back and forth between animation and effects, which would've been too expensive.

"Instead we went more linearly, where animation would animate the characters and the bridge and the gross movement, and it would go to effects, which would add such details as vibrations and breaking things apart. This was very crucial. The character group abstracted the controls for animation so it was simple to use. They kept the bridge moving very slowly. Once we rigged it, then we had to make sure all the characters were parented with the bridge. This was tricky because everything isn't locked on the bridge: you have cars that slide, dividers would break a bit because you can't really bend the bridge without causing tracks. Matte painting would also jump in to add fog to help sell the scale. The water underneath was also done by effects. Lighting water is always a challenge and in stereo it is a little more of a challenge because speculars and highlights have to match for both the left and right eye."

[18]

Cars and people stuck on the bridge had to react to the movement, as well as the turbulence from the fight itself. [18]

The digital effects supervisor concedes that there are a lot more vfx in MVA than any previous DreamWorks animated feature. "The scale is such that you're bringing down a whole building, so that affects the quality, too. Ordinarily we have elements of dust and breaking debris and add a crystal level of fine particle detail. But here the dust element would be cloudy and opaque because it's such a destructive effect. We had more model detail in the debris because each one is the size of a car."

He says some of the effects were done in Maya [19] and Vue for matte painting, set extensions and environments. Another off-the-shelf tool is Blast Code for fracturing and simulating objects.

Which brings us to the all-important 3-D. As nearly everyone knows by now, MVA marks the first all-stereoscopic production [20] at DreamWorks, and served as a test case for everything else yet to come.

[21]

3-D had to be managed so as to not bring viewers out of the movie, using a Blend tool to reduce eye strain. [21]

According to director Vernon, they were very cognizant about the negative aspects of 3-D, "which were strain, too much poking out at you and it was just a gimmick to get some [extra money out of viewers]. We wanted to sit down and figure out how we could overcome all of that, so we tread very lightly into 3-D. We said, 'OK, we're not going to stick things out at the audience, we're not going to pull anything off the screen unnecessarily, let's just look at the camerawork, look at the scene and see how it plays. Once we watched the whole scene and saw how the 3-D was acting, we realized it was controllable. 'Let's pull this character back a little bit.' And this was happening over two months of testing the boundaries. Once we really got comfortable with it, then we could literally pinpoint areas where we could reach off the screen. I've said this before, but one of the nicest things we heard about the 3-D was how beautiful it was at first and how much they forgot they were watching a 3-D movie once they really got into the story. Once Gallaxhar comes down to abduct Susan, and shoots out of the screen, suddenly you're reminded emotionally that this is in 3-D.

"They said we couldn't cut fast in the beginning. But we said we're making a hard-core action film. We have to be able to cut fast. We sent the technology team away and they came up with the Blend tool [to address that problem]. The reason you can't fast cut is if you have something really deep and then you cut to something coming out of the screen, you get that eye strain. So the Blend tool guides your eye from shot to shot during the cut and reduces the strain. We could cut fast and shallowed out the shots when they were on screen for 1.5 seconds.

"What was interesting about the test audience reaction [and the desire for more 3-D] was that Jeffrey [Katzenberg] told us to go in and break it. So we went in and found 10 or 12 places where we could really push the 3-D and out of that, four or five of them threw you right out of the movie. So we cut back and went back to the old version. But there were also a couple of places where it worked really great like Insecto shooting silk out at the camera, so the angle was changed."

Burgess points out some other subtle examples, too: "The shot where Susan has just grown to her new height and she's been drugged and falls onto the ground and reaches forward toward Derek with her hand. That's a shot that we lengthened and re-lensed to make sure that the hand was able to just hang out over the audience.

"There were a few places where we added shots to move the camera above things. For example, when the Quantonium [energy] is being sucked out of Susan in the extraction chamber and we wanted the goo to come from her up into camera."

[22]

The vfx team had to deal with a complex software and hardware setup to create the over-the-top visuals, with the added complication of 3-D. [22]

At first, Ramasubramanian says it felt more like a hindrance than a creative tool. "But then there was a moment when we went to see an early explosion shot in the dailies room and everyone suddenly experienced a palpable difference and realized that the extra work was worth the effort. There were growing pains of learning how to use the stereo hardware (in partnership with HP [23] and Intel) and setups at desks and software issues (in partnership with Autodesk using Maya) where you need to understand two camera techniques and then other techniques that had to be figured out later on. So if you have an object that is very reflective, for example, B.O.B, if the normals are changing too fast, it may be accurate in real life, but we can't leave it at that: we have to manage the left and right eye. We came up with tools where you could adjust how much depth there is in the reflection... we controlled the intraocular separation when computing reflections and refractions on B.O.B.

"Another thing was motion blur. If things blurred too much during an explosion, they became too transparent and you weren't able to place them in depth accurately. So we had to come up with ways of cheating the motion blur where we didn't let the streak get too transparent and preserved some of the opacity. There is a creative aspect to the whole thing. You have to decide what the stereo storyline is. But we found our sensibility through the process."

Bill Desowitz is senior editor of VFXWorld and AWN.