A New Chapter in 3D
Issue: Volume: 28 Issue: 12 (December 2005)

A New Chapter in 3D

You don’t have to read between the lines to notice that Mr. Finnegan’s Giving Chest, a just-released children’s Christmas storybook, is somehow different. While the base plot has a familiar ring with its uplifting holiday message about a person who discovers the joys of giving, it is the illustrations that immediately catch your eye. They are bold, dazzling, detailed...and 3D.

The fact that the book was crafted using CGI is not surprising given the author’s identity: Dan Farr, president of Daz Productions, a developer of 3D software and ready-to-use models. Another unlikely but familiar face behind the project is famed actor Dick Van Dyke, who lent his likeness to the lead character, toymaker Mr. Finnegan, and his voice to the words on the pages (he narrates the story on a CD that accompanies the title).

The spark for this project began shortly after Farr had met Van Dyke at the NAB conference in 1998. Not just a fan of 3D imagery but a CG artist himself, Van Dyke has been spotted at SIGGRAPH and other trade shows perusing the industry’s latest wares. “We built a relationship,” recounts Farr, “and one day I suggested doing a 3D animated project together, and he said, ‘Sure. What?’”

Far more difficult was coming up with the right project. Several writers pitched story ideas, and Farr even penned a script himself, but none of the concepts received a green light from the actor or his agents. Then, one concept struck a chord: Mr. Finnegan’s Giving Chest, a revised version of Farr’s original tale.

Farr’s intent for the project was threefold. In addition to the collaboration with Van Dyke, Farr wanted to show off the capabilities of the Daz products, and he wanted to tell a story that he had been toying with for some time. However, rather than try to produce an animation project immediately, Farr thought it would be more prudent to do a children’s illustrated book first and craft the visuals in 3D.

Between 2001 and 2002, Farr, along with his business partner, Chris Creek, and his staff, worked on scene development and created two short animations containing Mr. Finnegan and some of the characters in his toy shop. (The crew gave Van Dyke a copy of his 3D alter ego, which the actor animated and displayed during last year’s The Dick Van Dyke Show Revisited special on CBS.) “Then, we turned our attention to the story, alternating between it and the imagery,” says Farr.

Last Christmas, Creek finished sketching all the scenes and inserted the pencil and paper drawings, along with the story lines, into a booklet, thereby helping the group to visualize the project. In April of this year, the production really got moving when Farr and his team hired college intern Jacob Speirs to start composing the book scenes in 3D using Daz Studio, the company’s 3D animation package.

Although the fictitious Maggie is the focus of the story, more time was spent modeling Mr. Finnegan, mainly because his likeness was based on an actual person, and a well-known one at that. “We had to come up with a caricatured model that Van Dyke would be happy with,” notes Farr. “We wanted it to be flattering yet stylized and cartoony as opposed to photorealistic, which would have been easier to do with a digital scanner.”

To aid Creek in crafting the model, Van Dyke supplemented the pictures Daz took of him with photographs he had from years back, including some memorable pictures of him with Walt Disney and in scenes from Mary Poppins and Chitty Chitty Bang Bang, in which he portrayed characters that exemplified the traits and personality Daz wanted for Mr. Finnegan. These shots were then placed around the room while Creek sculpted a life-size clay model bust of Van Dyke.

In Mr. Finnegan’s Giving Chest, the snow globe handled by the toymaker contains a mini model of the town of Pinebough, built as a large, continuous 3D environment.

With an Immersion MicroScribe-3DX digitizer, the group scanned the clay model and imported the data into NewTek’s LightWave, where the crew cleaned up the mesh, remodeled the geometry, and then modeled the rest of the figure based on Creek’s initial sketches.

Once the Finnegan character was completed, Creek and his team modeled the rest of the cast and the scenery, creating the specific face morphs for these characters according to Creek’s sketches by using the hundreds of dials contained within each model.

Because the artists had carte blanche to shop at Daz’s online 3D model shop, they populated the scenes with as much prebuilt, ready-to-use content as possible. However, mostly they created unique models-including a mechanical dog named Switch, a heart-shaped diamond pendant, various toys, and more-to satisfy the specific needs of the script when it called for unique, esoteric objects. “We didn’t want to compromise and use something that didn’t quite match the story just because we had the model in our store,” explains Farr. Some of the models specific to the book are now on the store’s virtual shelves, and others (though not all of them) will be added incrementally.

Next, the artists rigged the models and added an Efrontier Poser skeleton, which was read by Daz Studio, the tool the team used for the character posing, lighting, and camera setup. With Right Hemisphere’s Deep Paint and Adobe’s Photoshop, the artists later textured the imagery and rendered it in Daz Studio’s 3Delight, a Pixar RenderMan-compliant renderer.

Because Farr wanted to have the book available in stores for this holiday season, the group was constantly challenged to complete the work in a timely manner. “If we needed an object that was unavailable in the store, someone at Daz would create it, rig it, and map it, and we’d just have to place it into the scene and pose it,” says Chad Smith, project lead. Yet, only a handful of Daz employees were dedicated regularly to the project, and even fewer worked on it full time.

The most complex scene in the book is Mr. Finnegan’s workshop, which comprises several million polygons. Each object contains high-resolution textures and is fully rigged, allowing the Daz team to pose every model in the picture.

As a result of the project’s staffing situation, meeting the publishing deadline often required both the artists to find creative work-arounds and the in-house programmers to write last-minute scripts. “Having the Daz programmers available to us at all times was invaluable in completing this project,” says Smith. “When we told them that we needed the software to do this or that, they had it done in a day or two, saving us large chunks of time.” Smith recalls an incident when an out-of-house employee needed to work on a scene over the weekend. This required Smith to manually locate, condense, and send him all the assets, which took hours to do. When Smith mentioned this process to programmer Rob Whisenant late one Friday night, by noon the next day, Whisenant had created a tool that, with a click of a button, automatically collected the assets and placed them in a zipped file-in a fraction of the time.

Indeed, having a wealth of 3D models at the artists’ fingertips and programmers at their beck and call provided untold advantages. Furthermore, using 3D tools to create the scene stills gave the illustrations a richer, more compelling look with a depth of field that is seldom seen in hand-drawn images-a look the artists never would have accomplished through traditional means. However, these advantages came at a cost. Because the prebuilt Daz models were prerigged and ready for lip sync, they were extremely dense. Although it was a time-consuming task, the team even rigged the models created specifically for the book so that the group could add the objects to the Daz store and also use them during the animation phase.

As a result, Mr. Finnegan’s workshop is filled with toys for all little girls and boys, and while the objects are stationary in the book, under the surface, they are ready, at the click of a mouse, to spring to animated life: Bears can be made to stand up and dance; toy unicorns can be commanded to gallop. And by using these rigged models, the artists could explore different camera angles and poses in an animation, and then choose just the right one for their still image on the page.

Smith recounts one time when the artists changed the rigging for Switch in the middle of production, to make the dog’s ears more floppy, its tail less polygonal, and its body stretch farther without distorting, all so they could generate additional poses for the stills. In another instance, they altered Maggie’s facial appearance after production began, to make the girl “a little cuter.” “When you look at a drawing on a piece of paper, you use your imagination 100 percent to picture it differently, whereas with CGI, you can brainstorm out loud in a visual way,” says Farr. “And you always find new ways of doing something better. Yet, if we hadn’t been working in 3D, those types of changes would’ve been far more difficult to make than just double-clicking on the existing scenes.”

By using 3D models that were animatable, the artists could picture each scene in a number of ways to see if the composition could be improved. After several iterations, Daz chose these poses of Maggie and Switch (top) and Maggie and her friend (bottom).

Despite this advantage, the group didn’t deviate much from the initial composition sketches, but there were times when the final image was radically changed after the artists reviewed the 3D scenes and discovered a better setup. (Daz included the initial sketches on the binding paper of the book.)

Looking ahead, Farr says the CG assets will come in handy not only for the animation, which will begin production after the holidays, but also for a series of books, which is another possibility. “Every time we would’ve had to redraw a complicated scene, such as Mr. Finnegan’s toy shop, our investment paid off a little more,” he notes.

The toy shop, the most complex scene in the book, comprises several million polygons-hardly surprising considering the high-resolution texture maps, dense geometry, complicated rigging with numerous body parts and parenting, and more. As Farr points out, you can zoom in on any figure and see the material properties and the color and bump maps. As the deadline loomed, however, the artists had no choice but to take some shortcuts.

Maggie, along with all the other characters, was modeled in LightWave and later textured within Deep Paint and Photoshop. Daz used its own 3Delight to render the images.

“It got to a point where if we could make the models easy to animate, that was a plus,” says Smith, who notes that there is a shot of an old woman who, from the side, looks loving and nice, and that is how she is shown in the illustration. However, if the model is rotated, the character looks hideous. “But that side of her face isn’t visible from the camera angle we used,” he says. “We’ll fix that later.”

And then there are the advantages to using CGI that are not quantifiable. “We know there is something that attracts people of all ages and demographics to 3D art and stylization,” says Smith. “But at the same time, we didn’t want the images to look too realistic or too cartoon-like. We were striving for a warm feel that you often don’t get with computer animation.”

Some of the book’s backgrounds are “flattened” 3D objects that are projected onto a hemispheric dome, thus giving them a dimensional feel that matches the rest of the environments and models only without the computational expense. For the most part though, the backgrounds are taken from snapshots of a single, large 3D environment. There are shops, a road, lots of people, a park with a lake, ice-skaters on the lake, and beyond that, trees, a church, and more buildings-all of which are high res.

And, there is snow. “We wanted to have snow everywhere, only the environment and the individual scenes were already huge without it,” says Smith. This left the group to ponder how it could reasonably blanket the town in white without it looking lumpy.

At first, the artists tried to use geometry, but the scene became too unwieldy. Instead, they modeled the snow and the road with individual pieces of geometry, and made the scene more economical by turning that geometry into two huge polygons-one for the road and one for the snow. Next, they used displacement mapping so that the bricks in the road poked up through the snow and the snow drifted against the sidewalks. Then, they added peaks and valleys, footprints, and anything else that would make the scene look realistic though with only a handful of polygons.

The falling snow presented yet another obstacle. “We realized we had to pull the snow out on a separate layer, but the snow alone took us a long time with our 3D render engine,” says Farr. Alternatively, the crew did the job in hardware, rendering the geometry planes with snowflake textures on them using the Nvidia 6800 and 7800 graphics cards in their homegrown PCs. As a result, each plane was rendered in 20 seconds compared to three hours in software.

“The Nvidia real-time performance meant we could tweak facial features, adjust lighting, and add textures easily,” says Creek.

Moreover, by using OpenGL to interact with and manipulate the scenes in real time, the group was able to produce many of its test renders in a matter of seconds. “The OpenGL produced such high-quality imagery with the Nvidia cards that we could see what we would get in real time as we laid out the scenes,” Smith explains. “We didn’t have to render several times throughout the day; what we got as our final render overnight was pretty much what we had expected.”

Adding snow to the 3D scenes was difficult but necessary for a holiday setting. Displacement maps give the surface snow a feeling of depth, while geometry planes with snowflake textures make the falling snow look light but still economical.

Smith notes that there are some scenes that are single layers, and those took longer to render-a process that was handled at night with the same PCs used to generate the imagery during the day. As the deadline neared and the group honed its pipeline, it got more efficient, reducing final render times from 36 hours to as few as four hours.

Will readers embrace this traditional story that is wrapped in untraditional packaging? Farr hopes so, especially given the time and expense that went into producing it. “In our case, we have an end goal of making an animation. But if we hadn’t, I doubt we could have gotten a publisher to support this simply because of the cost,” he says. Having Van Dyke’s support certainly helped Daz overcome that issue and secure a publisher and distributors (including Costco, Wal-Mart, and Barnes & Noble, in addition to online retailers, including Daz at www.givingchest.com).

“The Mr. Finnegan character is fictional, but the character really encompasses Van Dyke’s personality-he really is that fun, nice, and caring,” says Farr. “And when he is not working, he is in his shop tinkering around not with toys, but with 3D software.”

Immediately after the holidays, Daz plans to start preproduction of the animation, and hopes to have that project completed in two years. Currently, Farr is eyeing a 44-minute television special, but is considering a feature film. “We are rendering the scenes at HD resolution anyway, so it wouldn’t be a big deal to change our tack,” he points out. Looking to this next phase, Farr anticipates having to create more environments and additional characters. First, though, he will be looking for a seasoned scriptwriter to take the story to the next level.

Yet, even in its book format, the story line, like the accompanying illustrations, contains multiple layers that Farr believes will have different interpretations by those who read the book. As a result, this untraditionally illustrated children’s tale could turn out to be the next classic on everyone’s reading list.

Karen Moltenbrey is the executive editor at Computer Graphics World.