Creating images with AI — Resources
The most essential rules for building a prompt for a quality image, on a single page: structure, style, lighting, ratio and models. Save it or print it.
AI image — prompt rules
1Prompt structure
- Subject — the main object (who/what)
- Style — photorealistic, drawing, 3D...
- Composition — angle, shot
- Lighting + quality — at the end
2Style keywords
- photorealistic, watercolor, oil paint
- 3D render, anime, pixel-art
- An artist/style name adds precision
- Choose one main style
3Lighting and camera
- golden hour, soft light, neon
- bokeh — blurred background
- wide-angle, close-up, aerial
- 35mm, portrait lens
4Aspect ratios
- 1:1 — square (avatar, post)
- 16:9 — wide (banner, background)
- 9:16 — vertical (stories, phone)
- 4:5 — Instagram post
5Negative prompt
- State what should not be there
- For example:
blurry, extra fingers, text - Reduces low-quality results
- A separate field in many models
6Quality-boosting words
- highly detailed, sharp focus
- 8k, high resolution
- professional photography
- Don't overuse — in moderation
7Popular models
- Midjourney — artistic, high quality
- DALL·E — faithful to the prompt
- Stable Diffusion — open, flexible
- Leonardo / Firefly — free options
8Typical mistakes
- A too long, contradictory prompt
- Hands/fingers, text — may come out wrong
- Too many subjects at once — confusion
- Re-edit the result (iteration)