The set is a hollow, saturated, single-colour cube, and the prompt asks the model to randomise three things on each run: the colour, a dynamic pose and a dramatic wide-angle camera. That randomisation makes it a great prompt for generating a varied campaign series from one reference photo. The lighting notes are thoughtful too, asking for bounce light and reflections from the coloured walls, so the subject looks lit by the room. In our run, a fictional woman with a sleek bob braces against the walls of an orange box in a mid-leg-kick pose, wearing a black suit and white sneakers. To keep a series on-brand, replace 'randomly choose' with your fixed colour.
How to adapt it
Variable
What it changes
Examples
INPUT IMAGE
the person (identity preserved)
a consenting model's full-body photo, an AI-generated fictional person
bold single-color box
room colour (random by default)
tangerine orange, cobalt blue, acid green
dynamic pose
how the subject moves
bracing sideways against the walls, mid-leap, deep stretch
Tips
Use a full-body reference so the model knows proportions for extreme poses.
Fix the colour for brand consistency; let pose and angle stay random.
Keep 'tasteful clothing', which stops the styling drifting as poses get extreme.
