Creating 3D crowds on set for Wicked

Dimension became a part of the Wicked production team after shooting began and the team wanted to expand the crowds present in both Munchkinland and The Emerald City.

Using our digital crowds pipeline, we volumetrically captured the background actors - fully dressed in costume and make-up – processed the resulting data and handed it over to Framestore, the VFX team responsible for compositing these digital performances into the scenes.

For the project, we rolled out the Polymotion Truck which was built as a collaborative partnership between Dimension and MRMC. The truck is a mobile volumetric capture stage. For Wicked, our team deployed the stage to Elstree Studios. Over the course of two days, we captured 76 members of the supporting cast in 75 minutes of performances which totalled around 100,000 frames in 4K resolution.

The choreography and performances were carefully pre-planned before the talent entered the volumetric studio where they cycled through specific actions and reactions to ensure the resulting crowd animations aligned seamlessly with the scene’s narrative.

The stage inside Polymotion featured 106 cameras – 53 RGB and 53 infrared cameras. Ninety-six were positioned around the capture volume, with ten overhead, allowing for optimal results across varied performances. And to accommodate the wardrobe, which featured a lot of green, we repainted and covered the inside of the stage in blue screen material as opposed to the usual green to make sure the background wouldn’t interfere when keying out the footage.

We also used prop tracking to capture the brilliantly-designed hats worn by the citizens of Munchkinland and the Emerald City. During a capture session hats can shift, tilt, or bounce and without proper tracking, that subtle movement risks being lost or appearing disconnected in the final volumetric asset.

By treating the hats as separate props and tracking them with an Optitrack motion capture system, we accurately recorded their position and movement relative to each performer. This approach ensured cleaner reconstructions, higher visual fidelity, and greater flexibility for any post-production adjustments or replacements.

This set-up meant Dimension captured the performers as high-fidelity 3D assets directly on location eliminating the need for the cast to relocate or revisit the costume, hair, or make-up departments. The end result was a full 3D surface reconstruction of the actors with the backgrounds keyed out and cleaned up, delivered alongside a suite of tools to support the VFX workflow including motion blur, prop tracking and prop replacement. This meant the VFX team at Framestore could manage the assets and drop the individual performances where they needed to go.

Ozan Akgun is a producer at Dimension