AI prompt to describe a photo so you can recreate its look
Upload a photo to a chat AI that can read images and paste the prompt below. You get a seven-part description of how the photo looks (subject, clothing, setting, light, camera look, color and anything hard to recreate) and three short prompts that carry the same light, lens and colors to a new subject. Reuse those lines word for word in every prompt in a set, and the images should look more alike.
Fill in and copy the promptCopy the prompt
Your values are marked in colour.
Look at the attached image. I will use your description for a set of five team headshots with the same look. Describe only what you can see. Do not guess names, ages, jobs or personal details about any person. Return these sections, each in one to three plain sentences: 1. Subject and pose 2. Clothing and props 3. Setting and background 4. Light: direction, softness, color, and where the shadows fall 5. Camera look: framing, lens feel, depth of field, angle 6. Color palette and overall mood 7. Anything that would be hard to recreate, and why Then write three short image prompts that would recreate this look for a different subject. Each prompt must mention the light, the lens look and the palette. Label them A, B and C.
How to use it
- 1Upload the reference image to a chat AI that can read images.
- 2Fill in what you will use the description for, and paste the prompt.
- 3Read the seven sections and correct anything that is wrong. You know the photo better than the model does.
- 4Save the light, lens and palette (main colors) lines. Paste them word for word into every later prompt in the set, and change only the subject.
A good result looks like
A short, accurate breakdown of the look, in words you could hand to a photographer, plus three prompts that carry the same light, lens and palette.
Stay safe
- Do not upload photos of other people without their agreement.
- The model may describe a person wrongly. Do not treat its words as fact about them.
- If the photo is someone else’s work, use it to learn the look. Do not pass off a recreated image as theirs, and check the license before you reuse the photo itself.
If it goes wrong
4 known problems, each with a fix
- 01
Problem: The model guesses details about the person.
Fix: Repeat "describe only what you can see" and delete the guesses.
- 02
Problem: The light description is vague.
Fix: Ask a follow-up, "Where is the main light, and is it hard or soft?"
- 03
Problem: The prompts are too long.
Fix: Ask for "each prompt under 60 words".
- 04
Problem: It uses vague style words such as "cinematic" or "moody".
Fix: Ask it to replace each style word with what it can see, such as the light direction and the main colors.
Has it been tested?
Not yet. This one was written for this page and I have not run it on a specific model yet. Results will vary by tool, so run it once, check the result, and use the fix table above as your checklist. When I test it, the model, the date and the result will be listed here.
Questions
Why not just upload the reference to the image tool?
Some tools do copy a style from a reference image, but the result is hard to control. Words give you something you can edit, reuse and keep consistent across a whole set.
How do I keep a set of AI images looking consistent?
Copy the light, lens and color lines from the description into every prompt, word for word, and change only the subject. Rewording those lines between images is one common reason a set drifts apart.
Which AI can read an image and describe it?
You need a chat AI that accepts image uploads. Many do, but support depends on the product and plan, so check yours. I have not run this prompt on any specific model.
Build this for you
Want this built into your own workflow?
Tell me what you do by hand today. I will say what is safe to automate, what needs a person, and roughly what it would take. I reply within 24 hours.