Point cloud reconstruction from Google Jump stereo capture

Volumetric Video — VR Playhouse

Google Jump

Client · Google NukeX · Cara VR · Ocula · Houdini Volumetric Video for VR

The pure visual beauty of point clouds

Volumetric video for virtual reality — built with Google Jump, Houdini, and Nuke at VR Playhouse. I love the visual aesthetic of point clouds and their ability to represent and recreate three-dimensional space. I used to bake out final gather points from Mental Ray, combining them to re-create a static visual cue of an entire animation sequence and using them as combined elements for lighting calculations. With Houdini, working with point clouds is just so much fun.

Point clouds are now everywhere — for storing values, geometric reconstruction, and volumetric video. The Google Jump Assembler delivers depth maps at 8K, which lets us bring more advanced film techniques into virtual reality production and begin building entirely new production workflows for 3D reconstruction and more immersive experiences.

Stereo VR is not "light-field" — stereo panoramas won't represent correct view-dependent light changes — but it allows for greater depth and parallax, and for better integration with computer-generated images. Rectifying shot stereo with CGI is very tricky and inherently has problems. In these examples I'm using the Foundry's NukeX, Cara VR, and Ocula on the front end, and Houdini from Side Effects on the back end.

StudioVR Playhouse
ClientGoogle
DisciplinesVolumetric Video · Houdini
Front EndThe Foundry NukeX · Cara VR · Ocula
Back EndSideFX Houdini

The workflow included

01

The Shoot

Google Jump · Stereo Capture

Capture on the Google Jump rig — the starting point for a stereo VR pipeline where every frame will later be stitched, rectified, and reconstructed in depth.

Google Jump camera rig on set Google Jump camera rig on set Google Jump shoot in progress
02

Stitch & Depth

NukeX · Cara VR · Ocula

The front end of the pipeline: stereo stitching in Cara VR with NukeX and Ocula, turning the Jump capture into stereo panoramas — and its 8K depth maps into point clouds.

Rectifying shot stereo against CGI is where the hard problems live; solving it here is what makes clean integration possible downstream.

Point cloud generated from Cara VR stereo stitch
Stereo nadir views from the Cara VR stitch

"With Houdini, working with point clouds is just so much fun."

03

Switzer Falls

Houdini · Point-Cloud Reconstruction

The back end: Houdini rebuilding a location — Switzer Falls — as three-dimensional space from the Jump depth data.

This is 3D reconstruction toward volumetric video: greater depth, real parallax, and a path to more immersive experiences.

Switzer Falls point cloud reconstruction
Switzer Falls point cloud reconstruction
Switzer Falls VR reconstruction, wide view
Switzer Falls point cloud detail