Issue: Volume: 25 Issue: 11 (November 2002)

An Enchanted Village


Ride films, given their reliance on unrestricted camera motion, freedom of character movement, spectacular imagery, and stereoscopic creativity, have benefited greatly from recent advances in computer graphics. Yet, the combining of digital animation with a motion base theater has evolved into a somewhat formulaic experience, often set aboard a vehicle that can be easily simulated by the motion platform, such as a submarine or an airplane. The new large-format, stereoscopic 3D ride film Corkscrew Hill at Busch Gardens theme park in Williamsburg, Virginia, is breaking that mold with a bold narrative concept that uses its cast of unusual characters, rather than a vehicle, to drive the ride experience.

Written and directed by Jeff Kleiser and Diana Walczak, the film uses 3D effects to seemingly shrink the audience to the size of "wee fairies" and place them inside a magical wooden box. They are then whisked away on a first-person adventure through an enchanted 18th-century Ireland. The entire four-and-a-half minute experience unfolds from the point of view of the audience inside the box as it is handed from one character to another during the course of the ride. This innovative concept of literally putting the audience in the characters' hands required the use of a continuous camera shot traversing the film's 11 locations, since cuts would have disoriented the audience as the camera passed between the characters. Constructing the multi-location film with a single camera shot and a mechanical motion base designed to simulate non-mechanical animal and human movements presented the team at Kleiser-Walczak with a raft of challenges, in addition to those resulting from the film's stereoscopic presentation and bewitching setting.

From the confines of a magical box, viewers of the 3D stereoscopic ride film Corkscrew Hill are carried on a first-person adventure through a mythical 18th-century Ireland.




Corkscrew Hill's fantastical journey begins when two young boys, Paddy and Fenton, discover a small wooden box on a moonlit beach. When they look inside, they discover the audience, strapped into little seats. With the box in their hands, the boys run to a pub, where Duncan, the bartender, instructs them to find the "pooka," a magical black steed residing in the forest. The horse decides that the "fairies" are to be taken to Corkscrew Hill and sold to a witch named Moll. Leaping fallen trees and deep gorges, the pooka follows a cliff-hugging path as it carries Paddy and the box to the witch's castle atop the hill. Along the way, a troll named Hurley pounces from the trees, and banshees swoop down from a cemetery. Near the end of the ride, the witch's eagle-like griffin, Archibald, snatches the box, taking it high above the clouds, then drops it—and its contents—into a stomach-wrenching free fall to the ocean below.

The imagery in Corkscrew Hill has an illustrative, storybook feel, with lush greens awash in the deep violet-blue of a full moon and waning twilight, and interiors illuminated by the warm ambers of torches and oil lamps. This mythical setting enhanced the Old World ambiance of the quaint 18th-century buildings and furnishings (clocks, bar stools, lamps, signage, bottles, and so forth), which were meticulously researched for historical accuracy. The artists constructed the objects as NURBS surfaces in Alias|Wavefront's Maya, and textured them with Adobe Systems' Photoshop, Alias|Wavefront's StudioPaint, and Maya shading networks.

Among the most critical and challenging of the film's natural elements—including grass, leaves, water, and clouds—was the fog, which occluded the distant forest in a thick haze and gently misted the cemetery. The artists were unable to achieve a seamless transition between the fog levels of the two environments with the typical approach of rendering the scenes with Maya's environmental fog in a single color pass. That's because the leaves, created with Maya's Paint Effects, appeared to be rendered in unevenly gradated fog densities. Alternatively, the group created "fake" fog by assigning a white shader to all the geometry and rendering the scenes in a black fog. This render yielded a "fake" Z channel, which the compositor then used over a color pass to add fog at the appropriate depth and density during transitions from different environments.

Creating this complex environment was made even more daunting by the continuous camera move, which required the artists to construct the scenes using a contiguous virtual set. "The boys had to move [as if they were in real space] as they went out the back door of the pub, through the little yard, through the forest, and onto the cliffs," says producer Molly Windover. When completed, the gigantic virtual set resembled a large island with a small town nestled at the edge of a peninsula, with 2.5 miles of shoreline surrounding it, and a mile of woods leading to the witch's castle.

Some regions of that digital world were modeled in greater detail than, for instance, the areas glimpsed from the aerial view at the end of the film. And, because of the stereoscopy, every element had to be constructed flawlessly in 3D, since the common 2D cheats of painting in details or painting out blemishes, such as tears in a patch network, could cause slight inconsistencies between the left- and right-eye images of a frame, thereby ruining the stereo effect.

The imagery in Corkscrew Hill has a storybook look as audiences encounter places and creatures from Celtic mythology. Yet the buildings, their furnishings, and the landscape are authentic, as the models are based on actual period and geographic data obtai




To achieve a similar painterly style for the cast, character designer Leonardo Quiles created orthographic drawings of the main characters that were used as references by the modelers. Next, Walczak sculpted character heads, as well as full bodies for the pooka, Hurley, and Archibald, using SensAble Technologies' touch-based FreeForm modeling system. The rest of the team resurfaced the voxel-based FreeForm models as NURBS patches, using Paraform's surfacing software. In Maya, the artists re-parameterized the patches, attaching several of them to produce a continuous, uniform NURBS mesh comprising approximately 11 patches.

Duncan, the two kids, and the troll shared the same basic facial topology, while Moll and an older gentleman named McTavish, who greets the guests during the ride's two-and-a-half minute pre-show, were outfitted with a slightly different and denser patch configuration that conformed to the contours of their aged appearance. Though it sometimes made the animation process especially time-consuming, the density of these meshes was necessitated by the 1900- by 1280-pixel resolution at which the characters were projected onto the giant screen, especially in the close-up shots.

Prior to rigging the faces for animation, the artists bound the head and neck of each character with a lattice deformer, which helped to smooth the deformation of the patches by the morph targets and the neck joints. The team then created approximately 25 facial morph targets for each of the main characters, and controlled them with Maya's Blend Shape tools. These morph targets were modeled on what the company calls "muscle metaphors," which depict individual facial muscles in their flexed state. By combining these muscle-based shapes, the animators could not only form the standard facial images for articulating sound, but also achieve a broad range in a character's performance through diverse, non-linear movements.

The base skeletal setup, developed by animator Simon Sherr and animation supervisor David Baas, was scalable to all the biped characters in the ride. It featured a full complement of forward and inverse kinematic (IK) switches in the arms and legs, and a suite of Maya Set Driven Keys to articulate the feet and fingers through a library of predefined movements. According to Sherr, the group avoided using automated functions, such as center-of-gravity compensation and spine compensation for hip movement, which "usurp the animator's control and lead to robotic motion."

Unlike the fully clothed bipeds, Hurley, the troll, required a more complex binding process to animate the anatomy of his exposed upper body and legs. To simulate muscle flexing and realistic chest deformation, the group used Maya's sculpt deformers to generate bulging tissue and Maya Set Driven Keys to link the curvature of Hurley's spine with abdominal Blend Shape targets.

Next, the artists created the different types of facial and body hair covering the bipeds and quadrupeds. Fenton's curly red hair was fashioned with geometry and painted maps, while the witch's long, ragged tresses and the bartender's combed-back hair were produced with Maya Paint Effects. Maya Fur was used to furnish the coat of the griffin and the pooka, as well as Hurley's scraggly body hair and McTavish's white beard and bushy brows.

Unlike the complex characters, the glowing banshees were created in two passes. The first was a color pass for their geometry and clothing, and the second was an "X-ray" pass, which utilized the "facing ratios" tool in the banshees' shading network to make their edges glow. The artists then transported the two passes into Alias|Wavefront's Composer, where they adjusted the opacity of the color and the X-ray passes until their glowing translucent bodies emerged.

While the continuous camera move did not directly affect the overall character models, it did influence the animation of the NURBS panels used for their clothing, which, in most instances, could be wrinkled either manually with creative weighting or dynamically with goal weight scripts. However, the artists relied on Maya Cloth for a one-minute continuous movement of Moll's draping dress—one of the longest ever because of the preclusion of cuts in the film. The team further enhanced the cloth with a worn and dirty look by creating texture maps in StudioPaint and then applying them to a Maya shading network.

Hurley the troll was one of the film's more complex character models. To simulate his muscle bulges and chest deformations, the artists used Maya's sculpt deformers and Set Driven Keys. Maya Fur was used to generate his scraggly body hair.




Some of the most creative character rigging techniques were reserved for the stereoscopic camera, which consisted of two virtual cameras—one for each eye—positioned side by side. Prior to rigging the camera, the team simulated the fairy's-eye view of the world through slight adjustments to the focal length and the interocular distance between the left- and right-eye camera. "We used only slightly exaggerated human proportions," says Sherr, because the actual view from that perspective would have produced especially elongated limbs. The animators then generated a representation of the theater's screen in their Maya windows, allowing them to see any objects that veered off screen when they approached the viewer.

As a surrogate for the wooden box carried by the characters, the camera not only had to reflect the point of view of the audience inside the box, but also mimic the movement of each carrier, whether the character was running, flying, or on horseback. To accomplish this, the group created character-specific camera rigs by parenting a camera to a modified version of each character's IK skeleton. When the box changed hands, the animators would switch from one character's camera rig to another, animating it according to the character's motion. Early efforts at rigging the camera to match the motion of the characters as they ran, flew, or rode with the box in their hands included using a set of IK legs attached to a dynamically driven bouncing arm holding the camera. Unfortunately, the camera inherited too much movement with this setup, which would have resulted in a turbulent, if not nauseating, ride experience. The most effective setup entailed parenting the camera to the hand of each character's IK skeleton, and adding dynamics, goal weights, and various character-specific adjustments.

The artists used various testing methods, such as a virtual simulation of the ride, to ensure the quality of stereo 3D effects such as this one, in which Moll the witch brandishes a fork.




Because it was impossible to perform a completely smooth one-frame switch between the two cameras, the final render had to be executed from a master camera that synthesized all these uniquely rigged character cameras into one smooth shot spanning the length of the ride. Doing so meant the animators had to resolve the rotational phase shifting that could occur when the world space angles of the master camera tried to merge with the local space angles of a skeleton-parented camera. The animation team circumvented the difficulties of rotational solving by constraining the master camera to each character's camera with a set of vectors and point constraints. This enabled the master camera to trace the path of the character cameras in a linear, "closest distance between two points" fashion, thus negating any possibility of rotational phase shifting during transitions from local to world space.

Kleiser-Walczak has subsequently applied this vector-based constraint system to its character rigging technique, enabling animators to link the control sets of any number of characters—including their limbs, spines, heads, IK switches, and weapons—to create complex character interactions with unprecedented facility. "Animations that used to be difficult, such as characters holding hands or picking each other up in a continuous shot, are now incredibly easy for us," says Sherr.

After Paddy and Fenton discover the wooden box, they peer inside and find what they believe are fairies, which are actually the viewers, whose visual perspective is distorted through the film's digital and stereo effects.




After the animators created the master camera, they rendered nearly 30 layers for each location at a resolution of 1900 by 1280 pixels on Hewlett-Packard Linux and NT workstations connected to SGI Origin servers. The team then composited the rendered layers in Composer.

Next, the team created a virtual simulation of the ride by testing the convergence of the right- and left-eye images from the various viewing angles within the 60-seat theater. This was done by connecting an InterSense InterTrax 30 motion-tracking system to Sony's Glasstron glasses, wherein a 3D model of the theater was displayed, complete with a screen mapped with low-res frames of the film. The group further tested the stereoscopic animation at full screen size using a pair of Barco DLP projectors and a high-definition video server. Between these two methods, the group was able to check the convergence of the images during all the main 3D moments, which could look fine on a computer monitor but, once projected at full size or viewed from an angle within the simulation, could fail at a particular depth.

Corkscrew Hill is Kleiser-Walczak's most ambitious ride film since The Amazing Adventures of Spider-Man, which in 1999 turned Universal Studios' Islands of Adventure theme park into the mecca for thrill ride fans. That attraction became the first ride to accomplish the seemingly impossible feat of synchronizing stereoscopic CG animation with the moving point of view of a high-speed roaming ride vehicle. Yet, Corkscrew Hill has succeeded in breaking new ground in its own way. "We spend a lot of time in the hands of the characters in Corkscrew Hill, interacting with them at close quarters, and sharing in and accompanying them on their adventure," says executive producer Patrick Mooney. "In Spider-Man, our interactions with the characters are fleeting and less immediate."

Moreover, by wedding the sensory experience of the ride with the actions of its young protagonist, Corkscrew Hill has infused the medium with more dramatic expression and more intimate character interaction (see "Being There," pg. 56). It has also conjured the kind of vicarious experience that was once the domain of traditional character-driven feature films.

Using high-density models, including the boy Fenton, sometimes made the animation process difficult. However, the high-resolution imagery was necessary because of the ride film's large screen.




Martin McEachern, a contributing editor for Computer Graphics World, can be reached at martin@globility.com.
All images © 2001 Busch Entertainment Corp. and courtesy Kleiser-Walczak.


TOOLS
Adobe Systems www.adobe.com
Alias|Wavefront www.aliaswavefront.com
Barco www.barco.com
Hewlett-Packard www.hp.com
InterSense www.intersense.com
Paraform www.paraform.com
SensAble Technologies www.sensable.com
SGI www.sgi.com
Back to Top

Printed from CGW.com
Computer Graphics World Inc. | www.cgw.com | (800) 280-6446

Copyright 2013