AI Photo Prompts
Every photo prompt guide on GPTPrompts.AI in one place, sorted by what you want to make and which tool you use, plus a base formula and the rules that keep real faces recognisable.
Last updated October 17, 2026
Affiliate link, we may earn a commission at no extra cost to you.
Find the right guide by goal
Each guide below is a full prompt set for one job. If you are not sure where to start, pick the closest goal, copy one prompt, and use the base formula further down to adapt it.
Photos of real people
Start here when the image must still look like you or someone who agreed to be in it.
Occasions and seasonal trends
Time-bound pages. Check the date on each before you rely on a trend.
Looks and styles
Use these once your subject is right and you want a specific visual treatment.
Editing an existing photo
For photos you will share as real, keep edits close to what a careful editor would do.
Find the right guide by tool
The same idea needs slightly different wording in each tool. ChatGPT and Gemini take conversational edits on an uploaded photo. Midjourney leans on references and parameters. These collections are written for each tool's habits.
The base photo prompt formula
Almost every prompt in the guides above is a version of this template. Fill each line with something a viewer could see. Delete a line you do not need rather than leaving it vague.
[If editing a real person: Keep [person]'s face, facial structure, skin tone, skin texture, hairline and apparent age exactly as in the uploaded photo. Do not slim, smooth or reshape.] Subject: [who or what, doing what] Scene: [where, time of day, two or three concrete details] Light: [direction and quality, for example soft window light from the left] Camera: [shot type and angle, for example eye-level half-body, shallow depth of field] Style: [photographic, film look, illustration style] Crop: [1:1, 4:5, 9:16 or 16:9, and any empty space for text] Text: ["exact words" or NONE] Avoid: [the one or two things that keep going wrong]
A filled example: Subject: a woman in her thirties reading on a window seat. Scene: a small flat in winter, rain on the glass, a mug on the sill. Light: soft grey daylight from the window on the right. Camera: eye-level half-body, shallow depth of field. Style: natural phone-camera photo. Crop: 4:5. Text: NONE. Avoid: glossy skin.
Rules for photos of real people
When a real face is involved, the prompt has two jobs: change what you asked for and protect everything else. The guides for pictures of yourself and couple photos each include a likeness lock that lists what must not change, such as facial structure, skin tone, texture, hairline and age, and explicitly forbids slimming, smoothing and lightening. Paste it at the top of every turn, not only the first.
Make one change per turn and keep the best version as your base. When a face drifts, go back to the original upload rather than patching the drifted image. Compare eyes, teeth, ears and jewellery at full size before you share.
Consent comes first. Upload another person's photo only if they agreed, and show them the result before posting. Do not present an AI scene as a real event, and take particular care with photos of children.
Where AI photos usually go wrong
| Symptom | Usual cause | Where to fix it |
|---|---|---|
| Face no longer looks like the person | Many changes at once, weak input photo | Pictures of yourself |
| Two faces look alike | Features blended between people | Couple photos |
| Plastic, overly smooth look | Flattering adjectives, no camera language | Realistic image prompts |
| Edit spread beyond the selected area | Imprecise selection | Photo editing prompts |
| Style looks random across a set | No fixed style block | Aesthetic photos |
AI photo prompts FAQ
What is a good AI photo prompt?
A good AI photo prompt names the subject, the scene, the light, the camera angle, the style and the crop in plain, visible terms, and says what must not change if a real person is involved. It asks for one main change at a time. Vague words such as stunning, aesthetic or perfect give the model little to work with and tend to push faces toward a generic average.
Which AI tool is best for photos of yourself?
ChatGPT and Gemini both let you upload a photo and describe an edit in chat, which suits pictures of yourself. Midjourney is strong on style but works differently, with references and parameters. Features and plan access change often, so try the same photo and the same likeness lock in the tools you have and keep the one whose results still look like you.
How do I keep my face the same in AI photos?
Upload a sharp, evenly lit, unfiltered photo; start every prompt with a likeness lock naming the features that must not change; make one change per turn; and restart from the original upload when the face drifts. No tool guarantees an exact likeness, so compare results at full size and reject the ones that are not you.
Can I make AI photos of other people?
Only with their consent, and not in ways that could be mistaken for real events. OpenAI's usage policies, for example, prohibit using someone's photorealistic likeness without consent in ways that could confuse authenticity. Children's photos deserve extra care: only parents or guardians should decide to upload them.
Why do AI photos have bad hands or garbled text?
Hands and small text are where image models still make visible mistakes, especially in crowded poses and long phrases. Choose simple poses, give exact short text in quotes, and use the area-selection editing tool where available to fix only the part that is wrong.
Should I label AI photos when I post them?
Label scene swaps, recreations and art versions as AI images, especially on dating, professional or verification profiles where people rely on photos being real. Some platforms have their own labelling rules for realistic AI media, so check the platform you post on.
Primary sources
Checked October 17, 2026. Tool features and plan access change; each linked guide notes its own check date.