Case study
DITUS · Philia
A virtual-production R&D engagement across live MetaHumans, mobile virtual cameras, and real-time experiences — with LG’s CES 2023 keynote in-camera VFX as one of the public outcomes.
The engagement
A virtual showroom, and the in-camera VFX for a CES keynote.
DITUS and Philia brought me on-site for a six-month Unreal Engine R&D engagement — digital twins on PC and VR, live digital-human control, motion capture, mobile virtual cameras, AWS Pixel Streaming and the production tools that tied the pipeline together.
The flagship public outcome was the in-camera VFX work for LG’s CES 2023 keynote — machine simulations and technical visualizations rendered live behind the on-set photography, with operating guidelines for the production crew. Everything else fed into a virtual showroom the team could author together, perform inside and stream to a browser.
What I built
Five pipelines, one cohesive stack.
- LG at CES 2023 — in-camera VFX. Machine simulations and technical visualizations rendered in-camera for LG’s CES 2023 video production, plus operating guidelines for the on-set crew.
- Live avatar control. A MetaHuman driven by live voice and facial input with offline-rendered variants for quality reference, a photo-scan flow that turns real people into MetaHumans, and a parallel facial-capture rig on a Daz3D character.
- Live capture and virtual cameras. Facial capture driving the MetaHuman in real time, body motion capture for full-body performance and ARKit-powered phone virtual cameras paired with a second phone for facial capture — framing and performance recorded in the same take.
- Multi-user authoring, AWS Pixel Streaming, VR. Concurrent editing inside a shared scene via Unreal’s Multi-User Editor, AWS Pixel Streaming for any-browser delivery and a VR build of the showroom for headset walkthroughs.
- Virtual showroom front-end. The visitor-facing surface of the showroom — entrance flow, character selector and a day-to-night lighting transition setting the mood.
Live avatars
Avatars, several ways.
The headline pipeline is a live MetaHuman driven from voice and facial input, shown first below. After it: offline-rendered variants of that same character used as a quality reference, a photo-scan flow that turns a real face into a controllable MetaHuman and a live facial-capture test on a separate Daz3D character as a parallel rig outside the MetaHuman pipeline.
Virtual camera
Direct the shot from a phone.
An ARKit-driven mobile virtual-camera workflow that turns a phone into the camera body inside the Unreal scene — the director walks through the virtual set the way they’d walk through a physical one. A second phone running facial capture lets framing and on-screen performance be captured in the same take.
Streaming & multi-user
Many hands on the same scene; the result in any browser.
Two production-side workflows that kept the team moving fast: Unreal’s Multi-User Editor for concurrent authoring inside a shared scene, and AWS Pixel Streaming so the finished experience opens in any browser with no install on the viewer’s end.
The virtual showroom
Entrance, characters, time of day — and a headset to walk it.
The visitor-facing layer of the showroom — the first room you land in, the menu where you pick who you want to be, the lighting system that takes the scene from midday to night and a VR rig to walk the whole thing in a headset.
Stack