|“Action and adventure” are not the first words that typically come to mind when remembering Grimm’s fairy tale “Rapunzel.” But then Rapunzel, the beautiful girl trapped in a tower who lets down her hair, had not yet met the directors and writers at Walt Disney Feature Animation.
“The classic story takes place in a tower, but it’s hard to bring a small story to the big screen,” says Nathan Greno, who directed the movie with Byron Howard. “We kept everything that people love about the original story and expanded the world.”
“Nathan and I said, wouldn’t it be great if it wasn’t just a small story,” Howard adds. “It could be about a girl in a room and a guy coming to visit, but we wanted scope and breadth, a great, huge adventure and smart, contemporary characters people could relate to. We wanted a story worthy of our crew.
As in the original story, the evil Mother Gothel locks Rapunzel in a tower. But, rather than a prince, this modern Rapunzel meets a dashing thief, Flynn Rider, who has escaped his pursuers by hiding in her tower. She takes control. The result, Disney promises, is a hair-raising adventure. Seventy feet of hair, that is. CG hair. A challenge definitely worthy of the crew.
“We wanted Rapunzel to be a dynamic character,” Howard says. “So, we wanted her hair to do dynamic things. She uses it as a bullwhip, as a tool. She grabs a glass. We wanted it to get wet and blow in the wind. When we told the crew, they turned a little pale at first. And then they said they’d try to do what we wanted. They never said no.”
And as if creating, combing, and untangling 70 feet of hair weren’t enough, the directors pushed the crew further. Mother Gothel strokes Rapunzel’s hair. Characters hug. “You’ll notice when you look at other CG animated movies that there is little interaction with hair because it’s so difficult to do,” Greno says. “Even brushing against a shoulder. Our crew completely raised the bar. It will be fresh and new to the audience.”
Visual effects supervisor Steve Goldberg led the crew that pushed the technical bar higher. “We’ve got everything going on in this film,” he says. “Cloth on cloth. Hair on cloth. Hair on skin. Hair to hair. We have a lot of character contact in this show; it was integral to establish relationships, so we said we’d give it a try.
Goldberg provides an example: “Mother Gothel, the woman who kidnapped Rapunzel as a baby and locked her in the tower, uses the power of Rapunzel’s hair to stay young. She has a passive-aggressive relationship with Rapunzel and hugs her a number of times, sometimes around her hair, sometimes underneath her hair. And, she has long sleeves. Traditionally, we shy away from characters hugging other characters, but the directors wanted to support the notion that she has a smothering, suffocating relationship in the way she would hug Rapunzel. So, we removed a lot of restrictions we may have had on earlier shows. It was very exciting.”
Early in pre-production, the team evaluated available hair-simulation tools and then decided to develop an in-house engine to achieve the degree of control they needed. “Kelly Ward created underlying elements we call dynamic wires,” Goldberg says. “Jesus Canal supervised the simulation setup and the technical animation.”
To help understand how real hair moves, the crew brought in a model with five-foot-long hair. “We had her walk, run, stand in front of a big fan, tie her hair in a hook,” Canal says. “We wanted to see how hair reacts, the weight, mass, and volume.”
Then, directing animator Glen Keane, who had been the guiding force behind the film for years, helped the team understand what he wanted the animated hair to do. “He told us what he wanted to see when we posed and animated the hair,” Canal says. “He didn’t want us to ever lay the hair out in a straight, boring line.”
The technical team’s goal was to be able to hit a button and have the hair look and perform as intended: not necessarily to look real, but to look believable and appealing. “It worked for close-ups, but for wide angles, when we see the whole length of the hair, it was not possible,” Canal says
For those shots, the animators would put the hair into an initial pose, run the animation to see how the hair performed, and then the hair technical directors would tweak the animation and add constraints that affected the final simulation.
Although Rapunzel’s 70 feet of hair has as many as 140,000 strands when rendered, the animators and TDs worked with a small number by comparison: 173 dynamic wires, which looked like tubes and stretched, if necessary, the full 70 feet. The simulation engine drove these tubes, not the 140,000 strands. The rendering engine interpolated the tubes to create the number of strands needed for a shot. “Thanks to Kelly [Ward], the curves (tubes) matched, with about 90 to 95 percent accuracy, the shape and volume of the hair once rendered,” Canal says.
Riggers attached the tubes, which acted much like guide hairs in a typical hair simulation system, to Rapunzel’s CG head. Each tube came with between 30 and 40 controls that the animators and hair TDs could use to position and perform the hair, but they could add more on the fly if needed. “Or, they could reduce the controls to five, if that was enough for a nice smooth sway,” Canal says. Additional tools could freeze any length of hair that didn’t need to move.
The crew could also group the tubes into clusters, with all the tubes in a group following one tube’s lead, and animate how much the clustered tubes would follow the leader. “We could blend between animation and physical simulation,” Canal says.
If Rapunzel grabbed her hair, for example, animators might place the hair in her hand in the appropriate frame, and then the simulation engine would take over. To ease that transition, animation and simulation, using settings supplied by the TDs, would control the movement for varying percentages of the tubes in the cluster.
Low-res geometry — cubes and cylinders — with the same volume as objects, such as chairs and tables, that Rapunzel interacted with became collision objects for the simulation engine. “Interaction is very expensive, so the simpler the geometry the better, and the faster the simulation,” Canal says. “She uses her hair as a pillow, to open her closet and drawers, as a weapon. It was challenging and fun. The hair TDs had to come up with new ideas and new constraints for how things interact with the hair.”
When Rapunzel’s hair interacted with her skirt, the crew would usually simulate her hair first, using the hair as a volume for the skirt simulation. “Sometimes, though, it was the other way around,” Canal says. “We’d simulate the skirt and use it as a collision object that we fit into the hair engine.” For cloth, the crew used an evolution of the cloth engine built for Bolt. “It worked very well,” Canal says. “We had complex garments with multiple layers in some case, but it was fast. It took only seconds per frame depending on how many layers of cloth we had.”
Goldberg describes a particularly difficult cloth simulation that the cloth team was able to handle: “It’s when Flynn and Rapunzel are in the woods and something frightens her. Glen Keane had her climb on Flynn, arms around his neck, legs around his torso. He drew this as a joke in dailies, but the animator went ahead and animated it in CG. The cloth team not only had to simulate her skirt, but they had to have it function in a way that was still family-friendly. It’s a wonderful shot.”
To affect the simulated motion of Rapunzel’s hair in this shot and all others, the TDs set various parameters. “The simulation engine Kelly Ward and the team built was very realistic, but sometimes those results were not what we would expect to see or like,” Canal says. If the hair felt too heavy, for example, the TDs would lighten the mass. Appeal was always more important than accuracy.
Parameters also controlled the interpolation; that is, the number of hairs instanced from the 173 tubes. “When we say 140,000 individual hairs, that’s for close-ups,” Canal says. “For other shots, we had less.”
Texture maps controlled the hair’s look, color, shininess, and so forth, but lighters could also affect the look using a separate set of parameters. When Rapunzel sings a special song, her hair has a magical glow. “Her hair is like the sun,” says Dave Goetz, art director. “There’s a sense of warmth about it and about her.”
The crew used similar techniques for the other characters’ hair, although the animators rarely need to be involved in setting poses. “Flynn had controls on his front hair so the animators could animate on top of the simulation,” Canal says. “But mostly we did the other hair, even the horse’s mane and tail, with simulation alone.”
For the other characters, the technical challenge was more subtle. “The hair was obviously the big thing, but the second goal was to make sure we raised the bar in CG animation,” Canal says. “And I think we did. Having Glen Keane giving us notes was awesome.”
Keane, who had animated such beloved Disney characters as Ariel in The Little Mermaid, the Beast in Beauty and the Beast, and the title roles of Aladdin, Pocahontas, and Tarzan helped the team achieve the classic Disney animation the directors loved. “Glen wanted the people to touch faces, scratch, things like that,” Canal says. “So, we took a deformer-based approach for the faces and bodies that simulates muscles. It isn’t a muscle system per se. We didn’t want a realistic look; we wanted a cartoony look, but one that’s fleshy and organic.”
Because the characters wear clothes, the crew put more resources into developing facial rigs within Autodesk’s Maya that would help the animators create subtle, emotional performances. “This is an intimate movie with lots of close-ups,” Canal says. “The animators requested micro controls. We added blendshapes to hit micro targets.”
Even so, Clay Kaytis, a supervising animator on Tangled, describes the rigs as “simple.” “Carlos did a great job of distilling the movement of the face,” Kaytis says. “We work in Maya with a system of foundation shapes driven by controls. It does a lot without a lot of input, which is good because John Kahrs, the other supervising animator, and I thought it was the hardest thing we’ve ever done.”
Rather than a pipeline that moved a character from modeling to rigging to animation, for the initial setup, the crew worked in character-based groups of three to five people comprised of modelers, riggers, and animators. “We all worked together in teams from the beginning,” Canal says. “It was an open environment where we could share and express opinions. It was great.”
The idea for character teams originated with Bolt (see “Back to the Future,” November 2008), and Kaytis, who animated Rhino on Bolt, agrees with Canal. “Each person on the team learned what the other department needed so well that we wouldn’t have to explain after a while,” Kaytis says. “It was a great process. It’s funny how the power of five people working on a problem is so much stronger than one person butting his head against the wall.”
That same spirit of collaboration extended into the animation process. “Historically at Disney, and this goes decades back, we’ve had character supervisors rather than animation supervisors,” Kaytis says. “But on this film, we tried a different approach. We put the main characters in the hands of the entire crew. It was scary.”
One reason for using this approach was that with the directors Greno (a former story artist) and Howard (a former animator), the directing animator Keane, and two supervising animators, there were already five people looking at the animators’ work. “So we didn’t add another gate,” Kaytis says. “We had the animators take their work straight to dailies.
In dailies, Keane would pencil in his suggestions on a Wacom Cintiq, a display he could draw on with a stylus. “Glen would draw over the animators’ scenes and push to make them better,” Howard says. “For those guys, it was like a year-long master class in animation.”
The first sequence the animation team worked on, the confrontation between Flynn Rider and Rapunzel in the tower, took two months. “Glen would draw so much over those frames,” Kaytis says. “What the eyelash shape should be. The cheeks in profile. We had to figure out how to do this animation, to have this girl walk around and feel for her. Everyone just chipped away.”
And, the “master class” continued. “Glen was in dailies every day,” Kaytis says. “He’d stop on a pose, lean forward, and draw over the animation. Change a head angle. Push Flynn’s leg when he was leaping. He’s an animation genius. Every day, it was like, ‘Oh my God, I can’t wait for Glen Keane to draw over my stuff.’ You look at movies from the 50s — Peter Pan, Lady and the Tramp — and you just know those are Disney films, but you don’t know how to make it happen. I’ve worked at Disney for 16 years, and I think this is the first CG film which has that look. It’s a huge leap. And a huge amount of that is because of Glen Keane.”
While Keane pushed poses to make them better by drawing on them, the directors took a different approach. “Byron and I would get up and act out the scenes,” Greno says. “Animation can get too broad and then it doesn’t feel real, especially in emotional areas. We created this world. We want it to be believable. So we’d speak of things that happen in our own lives. We wouldn’t say, ‘This character is about to cry.’ We’d describe visiting someone in the hospital. People connect with that and put it into their work. If you do the job right, these characters become so real. They become people.”
IT TAKES A VILLAGE
Kaytis even extended this sense of creating people, not just characters, to the crowds. “We did the crowds basically all by hand,” he says. “We were stubborn about this. We wanted them to be better than any crowds in any other movie. Effects helped us with 1,200 people in the plaza, but otherwise, we hand-animated them for the most part.”
Rather than populating the village with CG characters driven by motion cycles and crowd animation rules, Kaytis wanted to create a town with real people doing real things. So, he invented little stories for each of the 36 distinct villagers. “I plotted out everyone in town every day,” he says. “The guy buying fish is doing that because he works in a restaurant, and we’ll see him in the restaurant later. The mother and father with a son are from out of town. They’re on a holiday, enjoying the sights. We see the village through Rapunzel’s eyes, so it has to be immersive. CG robots wouldn’t sell it.” Similarly, the 21 thugs in the bar each had a different personality, body shape, and design.
As he planned the shots in the village, Kaytis worked within a set designed to be charming, with buildings not taller than one or two stories, and everything at a graspable human level, as Goetz describes it.
“The town was like a backstage set,” Kaytis says. “You could shoot in one direction and then turn the camera around and find new cubbyholes and details in the buildings. You definitely became intimate with it. You know the street corners. You know that this guy works in that building. And once you know all that, it makes your life easy. It became kind of a game in a way. Definitely more interesting than stamping five characters.”
Kaytis shared the fun by giving each animator little vignettes to do. A guy fishing from a bridge. A woman sweeping the porch. Someone picking apples. Three students sitting on the stone steps. “It was kind of like the Secret Santa,” Kaytis says, “a cool little break from production.”
If the village seems familiar as well as charming, one reason might be that the directors were, according to Goetz, inspired by Disneyland. “We took pictures of everything,” he says. “And, of course, Disneyland is inspired by European buildings, so we kind of went shopping for charm and appeal.”
And, as with the village, the directors wanted the environments to be cozy and charming, too. “We have very few scenes where you look out over a huge vista,” Goetz says. “We cut off sight lines by pushing trees and hills in the way.”
Even within the environment, the particulate matter — dust motes created with particle simulation within a light ray — helped give the space a tangible feeling. One of the largest effects, though, was a dam-breaking sequence. “Basically, we’re using Physbam libraries that we have implemented and integrated into our pipeline for water simulations,” Goldberg says.
And, not surprisingly, Rapunzel ends up underwater in a cavern after the dam bursts. “We put the hair curves through a flow field,” Goldberg says. “When her hair glows, they can see the way out. We were all so excited and thrilled when we saw those tests come through. They involved so much interaction between the departments.”
It would probably be a stretch to imagine that in trying to create a CG film that captured the feeling of Disney films from the ’40s and ’50s, the crew also captured some of the adventuresome spirit of those early days of feature animation. But, maybe not.
“With CG, you can do anything,” Howard says. “And, as a result, live-action films are becoming more dazzling and eye-popping. Animation has to keep up. It has to be as amazing as live action. The action sequence in the underground cavern, we think it’s the biggest one Disney has done. It looks spectacular. We threw down the gauntlet and the crew rose to the occasion. They wanted to do this. They want to make better films. And this film will drop jaws.”
Just as they kept the classic feel of the Disney movies from the ’40s and ’50s while bringing the pacing and jokes into the contemporary world, the directors managed to incorporate the feeling of creating classical animation within a 21st century, collaborative working environment and with 21st century effects.
“You have to understand,” Kaytis says. “This whole building. Our crew. This team is so tight now. It’s really the process — the process was the best process. To see this team grow and step into the challenge, I couldn’t be more proud of the team.”