How to Write AI Image Prompts That Work

A good AI image is mostly a good prompt. The model can render almost anything, but it can only render what you describe — so the difference between a flat result and a striking one usually comes down to a few well-chosen words. Here's a simple structure you can reuse every time, the words that move the needle most, and copy-ready examples.

The four-part structure

Order your prompt as subject → setting → style → lighting. Name who or what the image is of, then where it is, then the art style, then the light or mood. Compare "a dog" with "a golden retriever puppy sitting in tall grass, watercolour illustration, soft morning light" — the second gives the model a clear subject, a scene, a look and a mood, so it has far less room to guess.

The words that change the look most

Three groups of words do most of the heavy lifting:

• Style — "photorealistic", "cinematic film still", "flat vector", "3d render", "anime". This decides the whole medium.
• Camera & composition — "close-up", "wide shot", "low angle", "macro lens", "shot on 35mm film". These steer framing and realism.
• Light & mood — "golden hour", "soft studio light", "neon reflections", "harsh overcast". Light carries the emotion.

Add one distinctive descriptor — "minimalist", "iridescent", "moody" — and you'll usually land somewhere more interesting than the default.

Match the aspect ratio to the job

Square suits social posts and avatars, portrait suits phone wallpapers and posters, and wide suits banners and video thumbnails. Setting the right ratio before you generate saves a lot of awkward cropping later.

Iterate, don't restart

Each run is unique, so generate a few and keep the best. If you like a composition but want to tweak the wording, lock the seed so the layout stays put while you change a detail — that turns "start over" into "fine-tune".

Ready-to-use examples

• Product: "A matte black cosmetic jar on a wet stone slab, soft rim light, deep emerald background, premium magazine advertisement, photorealistic"
• Portrait: "Close-up portrait, soft key light, neutral background, sharp focus on the eyes, natural skin tones, 35mm film look"
• Poster: "Vintage mid-century travel poster of a seaside town, flat screen-print style, grainy texture, bold title text, warm orange and teal"
• Thumbnail: "Bold thumbnail background, explosive orange and teal paint splash on the right, clean dark space on the left for text, high contrast, 16:9"

Drop any of these into the tool, swap in your own subject, and adjust one thing at a time. That habit — change one variable, regenerate, compare — is the fastest way to learn what each word does.

Try it in Quick Create →