What generative art is
The system’s rules decide which shapes, colours, movements and relationships are allowed, how they respond to input, and where the limits are. The output might be one evolving image or thousands of variations.
Four kinds of input:
- Rules and randomness. Algorithms that combine form, colour and motion within set bounds: consistent in character, never identical.
- Data. Live or historical: weather, traffic, scores, survey results, performance metrics, turned into shape, colour and movement.
- Interaction. The viewer’s cursor, touch, camera, microphone or device motion.
- Models. AI image, video and audio models directed by structured prompts and references, within the system’s rules.
Why it’s worth doing
Generative work does what hand-made work can’t, or can’t afford to:
- Scales. One well-designed system can produce every format, size and variation a campaign needs, and stay coherent.
- Stays current. A visual driven by live data changes when the data does.
- Responds. Work that reacts to the viewer holds attention and feels personal.
- Explains. Data shown through form and motion can be grasped intuitively, sometimes faster than a chart.
- Surprises. A good system produces results its designer didn’t plan, within limits they trust. That’s often where the memorable work comes from.
Where it shows up
- Website hero visuals that respond to cursor or touch.
- Campaign graphics in many variations from one creative system.
- Data visualizations people can read at a glance.
- Motion pieces and backgrounds that evolve instead of looping.
- Installations and large displays that react to the people in the room.
The role of the designer
Instead of making each image, the designer defines what makes one good here: the palette, the limits, how much variation is allowed before the work stops feeling like one family. They refine the rules until the system reliably produces work they’d put their name on. That’s art direction, applied to a system.
Taste matters more, because a system with poor rules produces poor work at scale.
How we build it
Our creative developers build it in three settings:
- In the browser, with custom shaders and real-time graphics that respond to input and run live on a website.
- In creative tools, where designers adjust parameters through simple controls and export finished graphics, no code needed.
- In pipelines, where data or prompts feed a system that produces assets for review.
We often start from 3D models or hand-made artwork. A Blender model can become a shader that responds to the mouse, turning a static hero into something you can draw across.
Keeping it performant and accessible
Shaders and real-time graphics can be demanding, so we design for a mid-range phone first, reduce detail where it won’t be noticed, pause off-screen animation and respect reduced-motion settings.
When a data-driven graphic carries meaning, the same meaning is available in plain text, for screen reader users and anyone who wants the numbers.
Testing the rules early
- Start with the idea the work needs to express and the place it will live.
- Prototype the system quickly, often in a day, to see if the rules produce anything interesting. First attempts often don’t, and with AI writing much of the prototype code, a failed rule set costs little.
- Tune the rules with you once something promising emerges.
- Build the controls or pipeline your team uses to keep producing the work after launch.
