About
[RE:INVENT / LV]A generative art installation created with our partner Production Club for AWS re:Invent in Las Vegas. A monumental LED cube that watches the crowd: real time AI reads each visitor's body and turns movement into living imagery across four visual worlds, inviting play first and spectatorship right behind.
AI THAT ANSWERS THE BODY
IR SENSING · SKELETON + DEPTH · LED
An infrared sensor extracts skeleton and depth from every body in front of the cube. Gestures drive the image and distance adds layers of control: the visitor is the brush, the cube is the canvas.
Custom development inside TouchDesigner feeds silhouettes, depth and skeletons to fine tuned diffusion models running on StreamDiffusion, generating up to fifteen high quality images per second, live.
How it works
[SENSE → GENERATE → PLAY]An infrared sensor captures every participant: skeleton data tracks the key joints while a depth image measures the distance to the screen.
Silhouette, depth and skeleton drive fine tuned AI image models through StreamDiffusion inside TouchDesigner: up to 15 images per second.
Effects and palettes recombine per world, so every pose returns a different answer. Proximity changes the picture; play is the interface.
The cube
[LIVE / WEBGL]On the floor
[LAS VEGAS]
The cube held the room: play first, spectatorship right behind.
