July 23, 2013, Siggraph, Anaheim, CA—The production of "OZ, the Great and Powerful" involved a lot of visual effects work to merge the live and CG characters. Visual effects supervisor, Scott Stokdyk from Sony Pictures Imageworks, along with Francisco de Jesus digital effects supervisor, and Troy Saliba, animation supervisor, described the various challenges involved.
Their environment includes effects and characters, so design resources included visual effects models and sketches, involving over 500 people at the peak. Pre-visualization was a necessity for this 3-D story, since the virtual world turned out to be about 3 times the size of Yellowstone National Park. To make matters worse, they could not use any pieces from the 1939 "Wizard of Oz" movie, so everything had to be original.
OZ The Great and Powerful - Sony
All of the design was done on the sets with the art department and the visual effects department. They developed a lot of concept art to help in set construction, which was mainly made from styrofoam. The sets all were blue screen to help with costs and size. Due to the extended shooting schedule on the set, a blue screen rather than green screen was used to minimize eye fatigue.
Lighting used the sun as the virtual main source so many scenes needed offsets from the set. Light design was in conjunction with 1k HDR images to capture all of the light and help to provide a setup for post. Everything was shot in 3-D with cinematic moves—booms and cranes—and not hand-held shots.
Virtual sets helped with the development and construction of the next set design, and also helped with edit and post visualization. One key was the improved ability to paint over parts of images to blend. They created many assets for hard surfaces, and also a digital nursery with trees, grass, flowers, and other nature objects to fill in scenery. Set dressing happened in the sets and in post. The nursery needed to reflect nature and project as clusters of objects.
They used layers to show movement for live action and to help with 3-D depth. To improve realism, they took pictures of real storms to get realistic clouds, but without the tornado funnels. They had to scale down the sets and the CG to get the wide 3-D shots for the storm scenes. They moved the cameras to capture a small window within the entire space.
In some sets, they built a real yellow brick road, but found that some scenes needed footings and a dirt path, so they rotoscoped out the people and replaces the bricks with dirt. Then they had to add the people back into the scene.
De Jesus was responsible for the effects animation. This work included procedural, simulations, physics, dynamics, weather including stormy skies, snow, wind, water, and waterfalls. They also had to create fireworks, explosions, smoke both natural and magical, and crowds. For the crowds, they key framed agents and used motion capture to set the movements.
The interactions between the action and life had to be invisible, so they rigged many of the assets for motion, etc. for all the shots. The environments had cloudscapes, ground fog, ash, and other moving objects. The magical people are associated with certain colors and effects: Theodora uses fire and is associated with red, Evanora used electricity and the color green, and Glinda has a rainbow and her colors are iridescent. Glinda has bubbles, a shimmery wall, and an impenetrable mist as her magic.
In the balloon crash scene, they had to make many takes and simulations to get the downdraft and river right. They used Houdini as their fluid solver, and Naiad as a plug-in for Houdini, for the high resolution work in the rapids. They found is easier to change the underlying surfaces to create shallows and randomization to change the rapids, rather than to try to change the water. They had to concatenate small segments to allow for smaller simulations.
Layering of objects effects, and animation over animation was very important. They clustered objects and partitioned the various objects with color codes for the different simulations of mist and spray, etc. The effects volumes of clouds, smoke, fog, fire, explosions, and magical mist were a mixture of rendering and effects.
Light is an issue as was interoperability with Arnold to set scatter and other optical effects. In cloud scenes, they had to create arches for the bubble ride, and angry clouds with flattened out tendrils and vortices for the wicked witch. The villain reveal was a fire tornado to slowly expose the wicked witch.
Troy Saliba noted that character animation was a real challenge. Two of the characters, Finley the flying monkey and China Doll, are all digital, but the people and images had to interact. Getting the concepts on to the screen took parts from pre- through post-production. Some of the issues were the digital characters' sizes and the need to interact with the people.
The voice actors were on stage for the rehearsals and improved the interactions on the set, but could not be on the set during filming. One solution was to make a puppet cam, a small display and camera on a boom, painted blue, to show the voice actor and provide a perspective of the character. This development allowed the actors to see each other during filming while the boom operator kept the eye lines correct. They also made a fixed camera and display for those scenes where they didn't have room for the boom.
They made a stuffed model of the monkey and a puppet for the China Doll. A puppeteer operated the puppet and changed their perception of the character as well as the effects so the movements and attitude of the puppet more closely matched those of the puppet. One of the unplanned efforts was to use a mixture of video and sketches as well as the physical models to help the actors see where the virtual characters were in the scene.
They used actual monkeys and baboons as models to get realistic motions and attitudes and increased their refinements through those studies. Annotations and paint over for the monkey and rigging helped capture the motions and fit them into their study models.
The digital characters needed a complex development process. The challenge was to make China Doll non-fleshy, but still have expressions. They hid the gross changes by making changes off camera, or through simplifed gross movements, sort of like Botox on the face. The difficult parts of the animation were fur, feathers, and cloth. The hard parts to bring to the capture was the live people and their reactions to the effects in the shoot.