Extra Celestial

The brief

Objective · Audience · Context

AI imagery collapses into sameness because most of it is made without anything to answer to. The question I set myself was whether a generated world could hold together the way a photographed one does — across a whole issue, not a single frame.

The idea

Creative concept

A civilization that has already arrived. Two atmospheric registers, three chapters, seven recurring artifacts, and one casting rule: she is not posed, she has arrived. The figure never performs for the camera. The world performs around her.

A figure in liquid silver facing the camera against a cobalt sea

$Iridescent beauty portrait — refracted light across pale skin$

Cobalt and Iridescent — two atmospheric registers, one accent per frame.

The Object — an iridescent mineral fragment suspended against a cobalt sky

Two speculative anchors — an Aman-coded property and a Mugler SS27 campaign. Neither commissioned.

Kept

Killed

Twenty-three images killed against seven rules. This one breaks two at once — a hypersaturated garment and a dynamic fashion pose, both on the never list. The hard kills are the good frames that break a rule.

Outcome

Issue 01 published and Issue 02 in production. The system ships as design tokens and a live site, so the world can be handed to another designer — or another model — without the look drifting.

Tools

Imagery: Midjourney v6.1 · ChatGPT
Motion: Runway
Editorial & site: Figma · Claude

Your system asks for reduced motion, so this site is running still. There's scroll animation here if you'd like to see it.

The brief

The idea

Visual direction

Applications

Selected process

Outcome

Tools