Research

R&D

We test new capture, rendering and AI tools by putting them into real productions. What works becomes part of the pipeline.

Published research

SMPTE Motion Imaging Journal

Cover of the SMPTE Motion Imaging Journal, October 2026, Vol. 135 No. 6, listing “Optimizing 180° Stereoscopic Cinematography for the Apple Vision Pro”

SMPTE Motion Imaging Journal · October 2026 · Vol. 135, No. 6

Optimizing 180° Stereoscopic Cinematography for the Apple Vision Pro

Stanley Lin, Nathan Fairchild, Spencer Lin, Akaash Tumuluri, Jordan Halsey, Scott Fisher · University of Southern California

Presented by Stanley Lin and Jordan Halsey at the SMPTE 2025 Media & Technology Summit.

The Apple Vision Pro shows more detail than any headset before it, so every flaw in immersive footage shows too. This paper sets out a repeatable, end-to-end pipeline for shooting 180° stereo film that holds up on it, using prosumer gear instead of a high-end immersive camera. The case study is the studio's VR short A Larger Living Space.

  • Capture. Canon R5 C with the RF 5.2 mm dual-fisheye lens at 8K, 59.94p, 12-bit Cinema RAW Light in Canon Log 3. 60p keeps motion smooth and reduces discomfort in the headset.
  • Exposure and lens. Shooting one stop over (f/5.6, exposed for f/8) cuts shadow noise. f/5.6 was the best trade between sharpness and light: f/2.8 was too soft, and f/7.1 was sharpest but too dark.
  • Lighting a 190° frame. The set was built without a ceiling and lit from the grid, with a CG ceiling added in post, and LEDs hidden in practicals. Light guides the viewer's eye without taking away their freedom to look around.
  • VFX. The lens was profiled with 3DEqualizer, photogrammetry and a Leica Disto survey; set extensions were built in Houdini with USD, Solaris, Karma and MaterialX, and composited in Nuke.
  • Finishing and delivery. AI denoising and a 16K upscale keep fine detail that traditional denoisers smear, before delivery as MV-HEVC with spatial metadata.

Research strands

What we're working on

Houdini and USD scene feeding an Unreal Engine render
Houdini + USD → Unreal
Real-time environment rendered in Unreal Engine
Real-time world-building
Unreal Engine VR environment from Perfect City
Perfect City · Unreal VR

Outputs

Talks, papers + releases