Talking Maleficent Pixie Facial Animation

The major tech advancement of Maleficent is Digital Domain’s facial capture of the three flower pixies played by Imelda Staunton, Lesley Manville, and Juno Temple, which required a more complex facial range and considerably more dialog than had previously been attempted. Video below produced by fxguide in partnership with Wired.

Carey Villegas served as VFX production supervisor and for DD: Kelly Port was VFX supervisor, Jonathan Litt was CG supervisor, and  Darren Hendler was Digital Effects supervisor.

But first, Paul Debevec, chief visual officer at USC’s Institute for Creative Technologies, utilized his famous Light Stage system to create digital 3D models of each actress. Staunton, Manville, and Temple went through a whole range of facial expressions and mouth shapes called visemes.

In terms of animation, character design entailed creating photo-real pixie versions of the actors, stylized but still making them recognizable. This also demanded extremely complex hairstyles, especially Temple’s, long free flowing curly hair. The hairstyles required extensive development of DD’s grooming and simulation tools. Additionally, there were complex, multi-layered dynamic wardrobes made of flowers petals, hairy thistles, leaves, and twigs.

Fortunately, Villegas and DD were able to leverage some new shader and rendering techniques doing development on Alex Proyas’ Paradise Lost before it was cancelled by Legendary Pictures.

“We were trying to take facial capture to the next level in terms of blood flow maps, compression of lips when blood drains out of them, and all those little nuances that get us away from this Uncanny Valley,” suggests Villegas. “And Maleficent gave us the opportunity to go beyond Alice in using the real actors’ faces and heads and digitally putting them onto 21″ size pixies and have them fly around like hummingbirds.”

Given that the pixies could be caricatured but with realistic skin shading, hair, and rendering made this a leap forward using DD’s latest four-camera system that’s HD and higher fidelity. Plus the FACS has been refined as well. In addition, as a replacement for Mova, they captured high-res facial shapes of each actor using a new system pioneered by Disney Research in Zurich, which had previously been utilized by Imagineering for the theme parks.

“We were able to build a high-density facial mesh for each of those [mouth] shapes but not only for moments in time but also in motion,” Villegas continues. “So now you can take that data set of 150 points from the camera and overlay [the facial mesh] on top of it to make sure that the interpolation is happening in the engine that Digital Domain has created and that it’s accurate to what the actor’s face is actually doing.”

Posted on by Bill Desowitz in 3-D, Animation, Below the Line, Clips, Movies, performance capture, previs, Tech, Virtual Production

Add a Comment