By Karen Moltenbrey
In the music video Rock DJ, when British pop singer Robbie Williams fails to attract the attention of aloof roller skating beauties and an attractive female disc jockey with his groovin' on the dance floor, he performs a dramatic disrobing. After shedding his clothes fails to impress the women, he reveals even more, eventually peeling away skin, flesh, and organs until his still-dancing digital skeleton finally catches the DJ's eyes.
To "pull off" the rocker's shocking striptease, animators could not miss a beat during the real Robbie's transition to a computer-generated model. "Our challenge was to create a seamless blend between the artist and his digital double, retaining his distinctive dance style accurately in the computer-animated content," says Damien Raymond-Barker, head of production at Clear, the London CGI house that created the digital dancer.
|British pop singer Robbie Williams shows that he's bad to the bone in an extremely revealing music video.|
Achieving this delicate transition required the Clear artists to match a prosthetic suit, muscle for muscle, on their digital model. Then they applied motion-capture data of the singer from AudioMotion (Banbury, UK) to match, step for step, his unique style of dance.
Creating the Hollow Man-like effects required using a combination of actual and digital means, according to Raymond-Barker. This was because the tight-fitting rubber suit that Robbie donned under his clothes, which enabled him to appear to peel away huge chunks of flesh, already contained a fair amount of bulk. If a prosthetic bone layer were to have been added, Robbie, who is fairly muscular himself, "would have ended up looking like Arnold Schwarzenegger in the suit." To eliminate some bulk, a CG model was substituted for the prosthetics whenever bone was revealed through muscle tissue, which occurred during the last 30 seconds of the 4-minute video.
To create the 3D skeleton encased in muscle, the animators purchased and modified a Viewpoint polygonal model of a human body's muscle structure rather than work from a cyber scan of the rock star, which Raymond-Barker says would have been more time-consuming. Using Side Effects Software's Houdini because of its "fast modeling abilities," the group traced each element of the Viewpoint model to create a NURBS surface before converting the file into an IGES format for export into Soft image.
The modelers then scaled the face and body elements in Softimage to match Robbie's exact proportions, using photographs and measurements of the singer for reference. "Getting the model to look like Robbie was difficult because there was no skin or hair to work with," says Raymond-Barker. "All the giveaway details were gone, so we had to ensure that his eye sockets, the nose region, the formation of his teeth, and other areas of his head and face matched his physical dimensions perfectly."
|To ensure that the 3D skeleton resembled Williams, the artists used his exact physical proportions to build the model, which was animated using motion-capture data from AudioMotion. |
For texturing, the director was seeking a look that resembled a "skinned rabbit," notes Raymond-Barker. At first, the team photographed various freshly cut pieces of raw meat, which were scanned and applied to the model. "But it looked too meaty," says Raymond-Barker. After examining other textures, the modelers decided instead to use scans of photographs taken while Robbie was wearing the prosthetic suit, and then filled in slight gaps using Adobe Systems' Photoshop.
Once the model was complete, the team also used Softimage to apply the lighting and motion-capture data. In an attempt to eliminate stretching and wobbling of the model when the motion was applied, the group created a kinematics subskeleton containing numerous secondary bones, to which the corresponding high-resolution 3D models were attached for easy weighting. "This allowed us to make minor cosmetic changes to the models and then simply reskin them," says John Harvey, head of 3D at Clear. "This enabled us to meet the tight deadline we faced for this project."
To capture Robbie's dance steps, the team at Audio Motion traveled to Bray Studios in Berkshire, UK, where the video was being shot. There, they dressed the singer in a Lycra bodysuit containing 44 reflective markers, which were captured by a nine-camera Vicon 8 high-resolution motion-capture system from Vicon Motion Systems during an hour-long session. This enabled the technicians to accurately capture Robbie's extremely energetic movements during his performance.
"The data captured by the cameras is highly detailed, and the final CG skeleton's movements are unmistakably Robbie's," says Chris Collinson, AudioMotion's mocap manager. "We captured the singer dancing and gyrating around a pole on the dance floor as only he can do. Traditional hand animation would have taken much longer and wouldn't have captured the subtle nuances of Robbie's movements." Toward the end of the live-action shoot, Robbie appeared to have gotten bored and casually glanced at his watch-an impromptu action that the AudioMotion group also captured and Clear later applied to the 3D model for inclusion in the video. "Anybody who knows Robbie will know these movements are from him."
Next, the team used its in-house Captivate mapping software to apply the motion data to a digital skeleton figure, also created with Captivate. Built according to Robbie's body measurements to match the 3D skeleton in the video, AudioMotion's stickman skeleton "drives" Clear's fully modeled 3D skeleton. Captivate's kinematic engine uses a mix of forward and inverse kinematics, so it eliminates "snapping" of the digital limbs. "That was important for capturing the fluid dance movement," says Rob Hughes, head of R&D at AudioMotion.
|Using its Captivate software, AudioMotion applied motion-capture data of a dancing Williams to a digital stick figure, which then drove the 3D skeleton model. |
To ensure that the digital Robbie skeleton fit into the high-pulse skating rink backdrop of the video, Clear used 3-Dfx.com's 3D-Equalizer to stabilize and track the live-action shots. Final rendering was done with Mental Images' Mental Ray, while Inferno was used for post production, especially to clean up the backgrounds and lighting in the CGI to match the actual set. "The biggest challenge was replacing the prosthetics with CG in such a fast-paced spot without it being noticed," says Raymond-Barker. "However, we could have used more time for compositing the longer shots."
Although Clear and AudioMotion put forth a tremendous effort on the project, the "revealing" video is receiving limited airtime. In fact, video footage of a recording session was incorporated into an alternative video for the song, so it could be shown on MTV, while the original video is being sold on DVD. But that's the way Robbie wanted it, notes Raymond-Barker. "He enjoys taking things to the edge, which is why people love him so much," he says. "And it resulted in a truly 'edgy' unique video."
Softimage, Softimage (www.softimage.com)
Houdini, Side Effects Software (www.sidefx.com)