Tech for Democracy

A machine trained on human hope and fear, asked to imagine tomorrow.

Tech for Democracy is a 30-metre immersive installation in which visitors type a single sentence about the future. A hope, a fear, a prediction and receive back an AI-generated image, a video of the generative process, and an email containing their result. Every submission joins a live, growing gallery on the installation walls.

We needed a system that could accept visitor input at the front of the installation, queue it for generation, produce an image and a corresponding video, and deliver both to the walls and to the visitor’s inbox without interruption. We built Fiddlehead, a Python backend, to manage this. Each submission receives a UUID, is validated for length and email format, and enters a persistent generation queue. The system connects to our image generation API over WebSocket, receives the generated frames, assembles them into a video using an interpolation step to smooth the latent traversal, saves the final image and video under the submission’s UUID, and sends an email to the visitor via SMTP. Originally built against VQGAN, we later migrated generation to our StyleGAN API for higher resolution and more controllable output; a configurable post-prompt is appended to every submission to steer the visual style without overriding what the visitor wrote.

The frontend is a React application with two modes: a prompt submission page where visitors enter their sentence, and an infinite-scroll gallery that displays every generation the installation has produced, in chronological order. A WebSocket connection to the backend pushes new results to the gallery as they complete, so the walls update in real time as new people contribute.

Tech for Democracy premiered at the Ministry of Foreign Affairs, Copenhagen, 2021. Subsequently shown at UN City Copenhagen, the New York High Line, and Biennale Nemo, Paris.

Commissioned by: Ministry of Foreign Affairs & Tech for Democracy