Gentle Monster Mock Campaign

Overview

A mock fashion-tech campaign blending Gentle Monster's avant-garde eyewear aesthetic with smart glasses technology. Filmed on a SCAD XR Stage using Unreal Engine with a physical set, under mentorship from Harbor Pictures, ETC, and Cat. Our concept explores AI vision as a third eye, revealing hidden layers of reality and bridging Eastern philosophy with technology.

As Virtual Production Lead, I handled technical pipeline setup, environment look development, material R&D, XR stage optimization, and on-set troubleshooting. I also contributed art direction, moodboards, and CG renders throughout production.

Concept & Art Direction

Gentle Monster doesn't make eyewear, they make wearable sculpture. Bold, architectural, unapologetic. We needed that same energy for smart glasses, but couldn't just make a tech demo. The concept we chased: AI vision as a third eye revealing hidden layers of reality.

After our pitch, mentors pushed toward K-pop visual language: rapid cuts, choreographed movement that highlights face and glasses, surreal elements like Gentle Monster's retail installations. Lighting direction was unanimous: hard, graphic stage lighting with strong shadows. Material guidance pointed to iridescent surfaces, mylar, and lens effects that make the glasses feel like they're revealing hidden layers.

Perforce & Color Pipeline

Set up the Perforce repository and Unreal Engine 5 project structure. Color pipeline was configured from day one with ACES workflow and OpenColorIO for consistency across CG renders, XR stage output, and final comp.

Master Material Systems

PBR UV Material

Standard surfaces: metal, glass, painted panels. Clean UV-based texturing with full PBR channel support.

Tri-Planar Material

Geometry without clean UVs. Projection-based texturing for organic and complex surfaces, more flexible for quick iteration.

RenderStream Setup

Configured RenderStream (RS2.0-UE5.5) with three camera channels: Outer Cam, Inner Cam, and AR. Each Cine Camera received a RenderStreamChannelDefinition component. On the AR camera, default visibility was set to Hidden with Atmosphere and Fog unchecked under general show flags.

Look Development

Environment: SaltFlats

Environment: Kinetic Museum

Pipeline Exploration: Point Cloud

Explored Gaussian splat to LiDAR PointCloud conversion using Apple's ml-sharp, converting 2D images to volumetric point clouds in UE5.

XR Stage Production

Load Test Optimization

Material Optimization

Converted select PBR materials to Tri-Planar for more flexible on-set iteration. Fixed missing materials and resolved Nanite artifacts on faceted geometry.

Scene Cleanup & Level Management

  • Deleted CineCameraActors in Levels (cleanup previs cameras)
  • Changed Geometry Brush Actors to Static Mesh
  • Deleted sub-level Post-Process Volumes, created Master PPV
  • Setup Master Level with Sequencer for managing level visibility
  • Aligned Pivot Point to (0,0,0)

Lighting & Rendering

Environment Optimization: FPS 95+

Shoot Day

On-Set Adjustments

  • Level_Lab applied 90 degree rotation due to root camera position
  • Roughness max 1.7
  • Fix reflection artifacts by setting Final Gather Quality to 4

Take Recorder

Recording camera data while shooting for post-production camera matchmove.

Technical Challenges

Level Sequencer failed to execute when simultaneously recording camera tracking data. After multiple failed attempts with Blueprint trigger boxes, Take Recorder settings, and record-before-play approaches, the final solution was pivoting to a per-level streaming strategy and removing positional keyframes, then slowing animation speed for a more realistic feel.

Color Pipeline Challenges

Color accuracy on set involved multiple factors: lighting cards on LED wall, physical lighting on set, RED Raptor sensor response & color science settings, and the color transfer chain (Unreal Engine → RenderStream → Disguise → LED Wall). It appeared blue-purple at first; adjusted parameters to achieve neutral color.

Post-Production & CG Renders

Continuity Quilt — Color Grade

Assembled a full shot continuity quilt after applying color corrections across all sequences, ensuring visual consistency from shot to shot.

End Shot CG Render

Rendered additional CG version of the end shot based on Sheng's animation, serving as a continuation of the closing sequence.

Breakdown

 

Credit

Team: Anna (Project Coordination), Bua (Storyboards, Set Dressing & Lighting), Sheng (Rigging & Animation), Robyn (Comp)
Model: Violetta
Makeup: Shannon
Camera Support: Ben, Julian, Deja
Technical Support: Conner, Aaron 
Mentorship: Harbor Pictures, ETC, Cat  

Image to 3DGS: Apple ml-sharp