How do you write good AI image prompts (Midjourney, DALL-E)?

You write a good AI image prompt by describing it in layers.

• subject
• then descriptors
• then style
• lighting
• composition
• finally the aspect ratio

Instead of "a dog," you say what dog, doing what, in what style, lit how, framed how. Specificity is the whole game — the more concrete detail you give, the closer the result lands to the picture in your head.

What structure should a prompt follow?

Build it in order, each part a short phrase:

• Subject — the main thing and what it's doing: "an elderly fisherman mending a net."
• Descriptors — key details: age, clothing, materials, colors, textures, expression.
• Style / medium — "oil painting," "35mm film photo," "3D render," "watercolor."
• Lighting — "golden hour," "soft window light," "dramatic side lighting."
• Composition — "close-up portrait," "wide establishing shot," "low angle."
• Aspect ratio — square, portrait, or widescreen, set by parameter or by asking.

What does a full prompt look like?

Putting it together: An elderly fisherman mending a net on a wooden dock, weathered face, wool sweater, 35mm film photograph, soft golden-hour light, close-up portrait, shallow depth of field, 3:2 aspect ratio. Notice every slot is filled with something concrete. Vague prompts get generic images; specific ones get intentional ones.

Does the same prompt work in every tool?

The structure carries across tools, but the dialect differs, so match the model you're using:

• Comma-separated descriptors — Midjourney-style tools reward short, high-signal phrases stacked together.
• Natural prose — DALL-E-style and conversational tools do well with a plain sentence describing the scene, and let you refine by chatting: "keep everything, change the light to sunset."

Whichever you use, treat your first prompt as a draft. The best images usually come after a few rounds — generate, see what's off, adjust one thing, and run it again. Iteration beats trying to nail it in one shot.