Generative AI Cube

The Gen AI Cube is an interactive art installation designed for the annual AWS re:Invent conference in Las Vegas. Powered by real-time AI-generated imagery.

Client: AWS · Disciplines: Events, Conferences, and Parties, Creative Technology, AI Development, 3D Animation, Event Production

A production by Production Club, a Los Angeles based creative studio.

The Gen AI Cube is an interactive art installation designed for the annual AWS re:Invent conference in Las Vegas. Powered by real-time AI-generated imagery, it recognizes human forms and reflects body movements in four unique visual languages: Explorer, Millennium, Flair, and Botanical. As guests step into the installation, their poses change instantly, inspiring both play and spectatorship.

Category One: Explorer

A rock formation in the desert with a moon in the background

A city street filled with lots of neon signs

A group of icebergs floating on top of a body of water

Modern multi-level building with curved balconies featuring abundant green plants

Category Two: Flair

A red vase sitting on top of a table next to other items

Red, yellow, and orange paint splattered on a black background

A purple flower with a gold center on a black background

A multicolored piece of art on a black background

Category Three: Millenium

A group of colorful sticks sticking out of the side of a blue wall

Abstract black modern sculpture with smooth curved shapes and holes

A colorful sculpture on a yellow background

Category Four: Botanical

A pink flower with green leaves on a black background

A bonsai tree in a pot on a black background

A green glass sculpture with a black background

How It Looked at re:Invent A room with a large display of a woman in motion

A large display of a flower in a building

A man standing in front of a crowd of people

A man standing in front of a large screen with icebergs

A man taking a picture of a large sculpture

A group of people standing in front of a large screen

A man and a little girl standing in front of a tv

A woman standing in front of a large screen

A group of people standing around a yellow box

A picture of a large object on display in a room

How the Cube Works

We used an infrared sensor to capture participants' body movements by extracting skeleton data and a depth image from the infrared sensor. The skeleton data then tracked key joint movements, while the depth image provided information on the participant's distance from the screen. This data drove the Generative AI visuals, with movements influencing the creation of dynamic imagery in real time. Powered by fine tuned AI image models, the system generated an impressive 15 high quality images per second.

A blurry image of a person standing in front of a sensor

Our custom development inside TouchDesigner allowed us to control human silhouettes, depth images, and skeletons, applying various effects and color palettes to create rich variations in the output. Gestures shaped the visuals, while proximity added layers of interactivity, allowing participants to control the evolution of the digital artwork through their body movements.

A.I. Photo Moments

A more lightweight version of the Gen AI Cube, this adaptation took personalization to the next level. Using real time face tracking and custom AI models, this installation instantly transformed participants into over 25 artistic styles, from Renaissance portraits to surrealist interpretations, creating unforgettable digital masterpieces.

People standing in front of a large screen

A large screen with a man's head made out of vegetables

A drawing of a man with his hands on his face

A painting of a man's face made up of different colored objects

A digital painting of a man's face and brain

A computer generated image of a human's brain

A sculpture of a man's face made out of wires

A woman's face with a red and blue stripe

A painting of a man wearing glasses and a red shirt

A woman with her eyes closed wearing a hoodie

A man's face made up of flowers and vines

Special thanks to dot simulate for creating the streamdiffusion TouchDesigner module which we used for the installation.