By Martin McEachern
Ford Motor Company took a giant step into the digital realm by creating several 3D models of various cars and trucks, including the Ranger (shown here), for use in regional dealer ad campaigns. The production of the virtual cars laid the groundwork for a digital model library to serve a range of promotional outlets, from television to the Web.
Although the computer-generated car has attained photorealism through advanced rendering techniques and the skill of digital artists, it is still rare for an automobile manufacturer to digitize more than two or three cars for its yearly lineup of advertisements. This is because most manufacturers still view the digital car as an expensive luxury for creating specific "beauty" shots and otherwise impossible imagery, not as a potentially cost-effective alternative to commercial photography, which remains the primary mode of generating marketing materials. So when artists at DKP Effects in Toronto were engaged by Ford Motor Company to digitize 14 of the company's 2001 vehicle models, it marked an extraordinary investment in digital as sets for a car manufacturer.
DKP, a 16-year industry veteran, had invested a year and a half of R&D perfecting both the realism and functionality of the CG car before the firm's achievements were recognized by Ford's national ad agency, J. Walter Thompson (JWT), through which the digital car project was produced. JWT and Ford were initially drawn to the undertaking to simplify the annual task of providing Ford dealers across the US with a selection of regional commercials that rival the aesthetics of the best national spots.
 |
|
Dealers typically are supplied with customized television spots featuring running and interior footage, as well as beauty shots, that are often composited with backgrounds that reflect a particular lifestyle or look familiar to local audiences. Filming these shots in an actual studio ordinarily requires a 20-foot diameter turntable to capture the car from a variety of angles, a 50-foot-high ceiling to accommodate the necessary lighting equipment, and at least 100 feet of space between the turntable and a camera track to get the shots. Reshoots are logistically unfeasible, confining the year's local advertising to the images produced during this single shoot. Hoping to eliminate the creative limitations imposed by this method, Ford opted to digitize its 2001 models to test the practicality and cost-efficiency of consigning these shots to the computer.
Within a strict time frame of 16 weeks, DKP digitized the vehicles and produced a total of 35 minutes of turntable-style ani-mation that was indistinguishable from a studio shoot. In addition, the company built a library of vehicle data that could be used to create low-resolution imagery for Web sites, mid-level-resolution animations for television commercials, and high-resolution images for bill boards.
In fact, the adaptability of the virtual vehicles to such a wide variety of marketing venues was one of the principal reasons Ford made such a large-scale commitment to the digital medium. According to Ted Rogers, a special effects producer at DKP, once the data has been collected and processed, the only issue is determining the resolution at which it needs to be output for each application.
Furthermore, major vehicle styling changes only occur about every five years, so Ford could lower future production costs by updating the 2001 digital models with any minor changes that occur the following year. "Car models generally don't suffer drastic design changes from year to year," says Rogers.
"A bumper, side-view mirror, or quarter panel might be refashioned. Using computer graphics technology, these alterations can occur quickly, without the added cost of a [studio] reshoot."
 |
| To create the virtual vehicle replicas for Ford, artists at DKP Effects set up their own assembly line. They digitized each vehicle and converted the resulting point cloud into polygonal and NURBS models. For texturing, they used shaders and texture maps |
Nonetheless, Rogers stresses that the ultimate strength of the virtual car lies in the creative freedom it affords a producer or director. For instance, the finite number of predetermined camera angles and lighting setups used during a turntable shoot offers limited flexibility for showing the cars from different perspectives and in different lighting conditions. "If a scene simply requires another camera angle to establish a different point of view," he says, "or if the lighting, color, or model options of a car needs to be changed, the re-take can be executed digitally at a minimum cost and without a reshoot."
While creating the digital footage, DKP ensured that the vehicles could be used to exploit these benefits-adaptability, reusability, and creative flexibility-by taking a modular approach to their construction and developing ever-expanding libraries of components, textures, lighting options, and camera setups.
The group began by scanning seven Ford models: the Escape XLT, Ranger XLT, Windstar SEL, Focus ZX3, F-150 Lariat SuperCab, Explorer EB, and Excursion XLT. Then the artists changed colors, badges, decals, and other trimmings to create their variant models. Initially, they attempted to use manual scans of the vehicle exteriors from Viewpoint DataLabs, whereby technicians placed adhesive tape in a grid pattern on the surface of each car, then used a MetreCom digitizing system from Faro Technologies, Inc. to register the cross points as 3D coordinates in the computer. Part way through the project, though, DKP realized the grid subdivisions were fine for generating sufficient point clouds for the distant shots but not for the close-ups. So the team switched to Capture 3D's ATOS (Advanced Topometric Optical Sensor) dual-optical 3D scanning system, which generated denser and more accurate point clouds.
The group then imported the scan data into Paraform's surfacing software and separated it into individual vehicle components, thereby reducing the size of the files without sacrificing detail. In Paraform, the animators converted the fractionalized point clouds into detailed polygonal meshes, and then patched all the minute holes, removed the defective polygons, reoriented the inverted polygon faces for proper texturing, and corrected all the other errors.
Once the meshes were airtight, the group re-created the polygonal components as a network of NURBS patches and profile and trim curves for use in Alias|Wavefront's Maya, where these features were lofted, trimmed, and merged. This process helped to create smooth and accurate surfaces from the polygonal meshes, and to reduce the size of the files so the components could be assembled in a workable vehicle master file. Wherever possible, trimmed and blended surfaces-which are memory taxing and highly fallible-were circumvented through effective use of Paraform's tangency tools at problematic seams.
 |
| Once the virtual vehicle model was assembled, DKP replicated the studio lighting setups and lensing techniques used during actual camera shoots, to achieve the same dramatic effects. |
Two modelers were assigned the task of hand-modeling elements common to all the vehicles, such as grill structures, mirrors, fenders, and bumpers, as well as layered components containing internal structures that the projected light couldn't reach, such as headlights, taillights, and signal lights.
Next, the artists assembled the original polygonal components to create lower resolution models for applications in which the NURBS meshes would be unnecessarily dense, hence slowing the rendering. Retaining both the NURBS and polygonal datasets would ensure that a functionally efficient car would be available for all applications, so that geometry only visible in a high-resolution medium would not, for example, en cumber the workflow of a Web project. After all the assembled models were visually compared with the original scan data and photographic references, they were furnished with a customized interior dataset from Viewpoint DataLabs, consisting primarily of seats and a dashboard.
For texturing the models, the artists used Adobe Systems' Photoshop as well as the layered shaders in Maya for creating sophisticated aggregate effects such as those present in the headlights and taillights. Figuring prominently in close shots, the complex physics of the lights would be the first to betray the disparity between reality and computer graphics. To create believable lights of manageable density, the artists used a composite of high-resolution bitmap files derived from scanned photos of the actual vehicles and a series of layered shaders to duplicate reflection and refraction as it occurs in moving car lights.
Special attention was also given to capturing the appearance of clear-coat paint. Senior animation artist Anna Wagner developed a shader to simulate the specularity and reflective quality of the surface, emulating the way it registers highlights and hot spots, and seems to reflect light at oblique angles, giving it a "wet" luster.
Finally, director of photography Steven Finestone, renowned for his studio car shoots, spent two weeks with Wagner, finding digital equivalents to the lighting setups for his turntable shoots. Typically, he stages these shots in airplane hangars to accommodate the enormous light boxes and reflecting panels required to cast a flattering pool of low-key "liquid light," which causes reflections to flow continuously over the car as the camera viewpoint changes. Using Maya, Wagner implemented sets of point lights to simulate light boxes and build a virtual equivalent of the hangar lighting grid. The size and angling of the reflecting panels, the intensity of the lights, as well as the focal length and other properties of the lens were also reproduced.
The entire project harnessed the combined efforts of 10 digital artists, who shared the modeling, texturing, lighting, and animation duties. When the project was completed, the ad agency lauded it as some of the most realistic digital car foot age they had seen. For DKP, the imagery brought to fruition the market research venture the company had initiated two years earlier.
DKP has also expanded its digital focus into environment simulation to help car manufacturers showcase their vehicles within their demographic domains, without the complications of traditional matte photography. This marketing convenience would allow them to quickly frame, for instance, an SUV in a mountainous terrain, a sports car on an open highway, or a luxury sedan in an elegant cityscape.
Using a digital palette, the artists can create photorealistic and illustrative environments that are fully animatable by using particle generators, such as Maya's Paint Effects, for dust, snow, rain, plant life, and other organic details; fluid simulators for water and mud; and photogrammetry for re-creating 3D models and camera angles from photographs. Likewise, they can generate low-resolution pseudo environments for Web sites and other memory-restrictive venues by mapping panoramas-created with RealViz's Image Modeler and Stitcher-to planes, tubes, or spheres in Maya or Side Effects Software's Houdini. DKP first implemented this technique on Toyota's Web site.
Employing these types of virtual settings, along with digital car models, has significantly lowered the turnaround time required to create vehicle advertisements for nearly every medium. "Complicated national commercials can now be designed, produced, and delivered in about eight to 12 weeks, dealer spots in about two weeks, and most Web applications within days," Rogers estimates.
By contrast, linear shoots are subject to a host of variables that demand much longer production schedules, including weather conditions, location restrictions, and prep time. However, a non-linear production can proceed without having the car on the set-often a necessity when the actual prototype is unavailable-and can incorporate camera movements that would be impossible or cost thousands of dollars to do on location. Moreover, the use of digital assets is allowing regional commercials, produced with limited budgets, to employ elaborate effects that were once reserved for national campaigns.
Still far from supplanting traditional photography and video, the use of virtual cars quietly continues to make the daunting logistics of mounting a massive, multimedia ad campaign more easily surmountable, and in time, may eventually join CAD as an indispensable ally of the automotive industry.
Martin McEachern is a writer and digital artist living in Toronto. He can be reached at
icetitan@hotmail.com.
Adobe Systems · www.adobe.com
Alias|Wavefront · www.aliaswavefront.com
Capture 3D · www.capture3d.com
Faro Technologies · www.faro.com
Paraform · www.paraform.com
RealViz · www.realviz.com
Side Effects Software · www.sidefx.com
Viewpoint DataLabs · www.viewpoint.com