◈UseAPrompt
🖼️ Category

AI Image Prompts

Detailed prompts for Midjourney, DALL-E, Stable Diffusion and other image generators — lighting, composition, and style dialed in.

Every great AI image starts as a sentence. Not a wish — a recipe: subject, light, lens, mood, medium. The prompts in this collection are written that way, which is why a cinematic portrait comes out looking like a film still instead of a stock photo, and a product shot looks like it came off a marble table in a studio, not out of a 3D render.

What makes an image prompt work

Image models don't understand "beautiful." They understand specifics: "Rembrandt lighting, catchlight in the left eye, 85mm lens" gives the model a thousand training images to agree on. Vague adjectives give it a thousand images to argue about. Every prompt here front-loads the decisions a photographer would make — light direction, depth of field, color palette, texture — so you start from a directed shot, not a lucky dip.

How to get consistent results

  • Change one variable at a time. Swap the lighting phrase, keep everything else. You'll learn what each phrase actually does within a few generations.
  • Set the aspect ratio on purpose. Portraits want 4:5 or 2:3, book covers 2:3, wallpapers 16:9. A great image in the wrong crop is a mediocre image.
  • Keep a style anchor. Reuse the same closing phrase ("shot on Kodak Portra 400, soft grain") across a set and the set will look like one shoot.
  • Iterate on near-misses. If the composition is right but the mood is off, don't re-roll from scratch — adjust the mood words only.

Chasing a specific look? The portrait prompts collect the people-focused shots, and the viral photo trends hub tracks the styles currently flooding Instagram and TikTok.

The range in this collection

Seven prompts, seven different crafts. The isometric room diorama is pure cozy geometry — a tiny world in a box, endlessly re-themeable. Double exposure portraits blend a face with a landscape until both mean more. The pixel-art sprite sheet speaks fluent game dev, the fantasy book cover knows that a cover's job is to be read at thumbnail size, and the minimalist architecture shot proves restraint is a style, not an absence of one. Work through the set and you'll feel the same grammar underneath all of them: subject, light, lens, mood — in that order.

8 ai image prompts prompts

Frequently asked questions

Which AI image tool do these prompts work with?

All of them, with minor dialect differences. Midjourney rewards comma-separated visual phrases, DALL-E and ChatGPT prefer full descriptive sentences, and Flux handles both. Each prompt here is written in descriptive language that ports across tools — the tested-on label on each page names where it was tuned.

Why do my AI portraits look generic?

Almost always the lighting. 'A portrait of a woman' hands the model total freedom, and freedom defaults to average. Specify the light (window light, golden hour, Rembrandt), the lens feel (85mm, shallow depth of field), and one texture (film grain, freckles, linen shirt) and the average disappears.

What aspect ratio should I use?

Match the destination: 4:5 for Instagram feed portraits, 9:16 for stories and Reels covers, 2:3 for book covers and posters, 16:9 for YouTube thumbnails and wallpapers. In Midjourney that's --ar 4:5; in ChatGPT just say 'vertical 4:5 composition' in the prompt.

Can I sell images I generate with these prompts?

On paid plans, Midjourney, DALL-E, and Flux all grant commercial rights to your generations — but check your specific plan's terms, and never generate trademarked characters or real people's likenesses for commercial use. The prompts are free to use however you like.

How do I fix weird hands and faces?

Generate at the highest resolution available, keep hands out of the focal point when you can ('hands in pockets', 'holding a cup with both hands' reads better), and use the tool's upscale or inpainting pass on the problem area instead of regenerating the whole image.

Official resources

Go deeper with the official docs and tools behind these prompts.