Case study / 06

Experimental Generative AI Art.

An underwater speculative world developed through generative image production, art direction and post‑production.

Tactical divers in an underwater scene from the experimental generative AI art series.

Project overview

This self-initiated study combines deep-sea operations with the visual language of space exploration. The project develops an underwater tactical world through recurring characters, equipment, lighting and atmosphere, building continuity across an extended sequence of generated imagery.

The art direction establishes a mineral green-black palette, dense particulate water, enclosed light sources, pressure-resistant machinery and recognisable crew suits. Prompt iteration in Midjourney and Adobe Firefly is followed by image selection, compositional correction and retouching, with each frame considered in relation to the surrounding sequence.

Portrait and landscape compositions alternate between crew formations, submersibles, industrial structures and wider operational scenes, alongside a motion sequence. The work explores how concept development, visual constraints, curation and post-production can shape generated material into a consistent narrative environment.

Role
Art direction · Generative image production · Retouching
Scope
Concept development · Visual world-building · Prompt iteration · Image selection · Retouching · Still and motion sequence
Credits
Self-initiated study · Created with Midjourney and Adobe Firefly

Design highlights

Visual continuity
Recurring pressure suits, submersibles and industrial equipment establish the visual vocabulary. Character appearance, machinery scale and atmosphere guide image selection throughout the sequence.
Light and colour
A mineral green-black palette, suspended particles and enclosed light sources connect underwater portraits, vehicles and expansive compositions through a shared cinematic atmosphere.
Image production
Prompt iteration is followed by selection, compositional correction and retouching, bringing individual generated images into the established visual rules of the wider narrative.

Your next project

Something along these lines?

Discuss a similar project