The anatomy of a good prompt: subject, style, light, and camera

When we analyze thousands of top-performing prompts across our tools, the same four-layer pattern shows up every time: subject, style, light, and camera. Understanding them lets you debug any mediocre generation.

Layer 1 — The subject

What or who appears, in as much detail as possible: age, expression, pose, wardrobe. The more specific you are, the less room for interpretation you leave the model.

Layer 2 — The style

Photography, illustration, watercolor, pixel art... Style determines which visual references the model draws on internally, and should always come after the subject, not before.

"Subject first, style second: the model decides WHAT to draw before it decides HOW to draw it."

Layers 3 and 4 — Light and camera

Light defines the mood (dramatic, soft, natural) and camera defines the framing (wide shot, close-up, low angle). Adding these two layers last is what turns a correct image into a memorable one.

Once you have the four layers down, our guide on how to write AI art prompts walks through complete examples.