Home / AI art and image generation
Fig. 04 — Image generationAI art and image generation
An image prompt describes a scene: subject, style, composition, light, material. Here is the structure that gives repeatable results.
Subject
Who or what, doing what, where: “a lighthouse on a Breton cliff at dawn”.
Style and medium
Photograph, watercolour, woodcut, 3D render, vintage poster: name one main medium.
Composition and light
Framing (wide shot, low angle), focal length, time of day, soft or hard light.
Materials and colours
Texture (grain, paper, brass) and a limited palette (3 hues) for coherence.
Parameters
Aspect ratio, seed, model version: note them to reproduce or vary an image.
Vintage poster of a Breton lighthouse at dawn, three-colour woodcut (midnight blue, ochre, cream), paper grain, vertical 2:3 framing, raking light, large empty area at the top for text.
2026 note: recent image models understand full sentences better than keyword lists; negative prompts remain mostly useful on Stable Diffusion-type models. Check usage rights and labelling duties for generated content.
- Midjourney — DocumentationParameters, styles, references.
- Stability AIStable Diffusion models and API.
- European Commission — AI ActTransparency rules for AI-generated content.