ChatGPT Photo Editing Prompts: 13 Ready-to-Use AI Prompts (2026)

AI photo editing means changing an existing photo by typing what you want instead of working through menus and sliders: brightening a dark shot, removing a stranger from the background, sharpening a blurry photo or putting a phone-shot product photo on a white background. As of September 2026, ChatGPT and Gemini do this on their free plans too, and the result depends less on the tool than on how clearly the prompt is written.

According to Google Keyword Planner, the Turkish query for 'AI photo editing' gets about 8,100 searches a month in Turkey, and that number rose to 9,900 in the last two months. Yet most lists in the search results offer style prompts that turn your photo into a Pixar or anime character. This guide is about fixing the photo you already have without breaking it: 13 copy-paste prompts, the limits of both tools checked against their official documentation, and the five locks that decide the result.

What Is AI Photo Editing and How Does It Work?

AI photo editing relies on an image model that reads the photo and the text prompt together, regenerates the requested area and leaves the rest alone. Google's Gemini API documentation describes this plainly: when you provide an image and describe the change, the model matches the original image's style, lighting and perspective. According to the same documentation, adding elements, removing elements, changing the style and color grading are all possible with a single prompt.

The difference from a classic editing app is that you don't have to draw the mask yourself. In Gemini you describe the area to change in words; Google calls this defining a mask conversationally, and its example prompt changes only the blue sofa while leaving the rest of the room as it is. In ChatGPT, you can mark the area with the selection tool in the image editor and write a prompt for that region only. Both methods share the same weakness. The model redraws the edited area, and if you don't list what to keep, faces, hands and text can drift too.

How to Edit Photos with AI: 5 Steps

AI photo editing takes five steps: pick the tool, upload the original photo, write the one change and the things to keep, compare the result with the original, and if needed keep going by fixing a single point. The flow is the same in ChatGPT and Gemini; only the way you select an area differs.

1. Pick the tool: Start with ChatGPT if you want to change a small area precisely, and with Gemini if you want to edit in several steps while keeping the person's likeness.
2. Upload the original: Add the full-resolution file from your phone, not a compressed copy from a messaging app or a screenshot.
3. Write the one change and what to keep: Describe what will change in measurable terms, and list what must stay the same item by item.
4. Check the result side by side: Zoom in and see whether the face, hands, edges, text and shadow direction match the original.
5. Fix one point at a time: Ask for the one thing you don't like in a new message; if the result is badly broken, start over from the original in a new chat.

13 Ready-to-Use AI Photo Editing Prompts (Copy and Paste)

The prompts below are written to work in both ChatGPT and Gemini. Attach the photo, replace the fields in square brackets to fit your photo, delete the brackets and send. Every prompt has the same skeleton: what changes, what stays, how the new element matches the scene, which side effects you don't want, and the output format.

Light, color and tone prompts

Edit the attached photo; do not generate a new image, improve this same photo.
Change: The photo is dark and grainy because it was taken [indoors / in the evening]. Raise the exposure by about one stop, bring back detail in the shadows and reduce noise.
Keep: Facial features, expression, hair, clothing, framing and every object in the background must stay exactly the same.
Match: The light should look natural, like soft daylight coming through a window, with no harsh flash look.
Limit: Do not lighten the skin tone, do not smooth the skin, do not add filters.
Output: A high-resolution photo with the same aspect ratio as the original.

Fix the colors in the attached photo.
Problem: Because of [yellow bulb light / fluorescent light] the photo has a [yellow / green] cast; the white wall and white clothes don't look white.
Do: Neutralize the white balance; whites should be truly white and skin tones should look natural and healthy.
Tone (optional): Then apply a [warm and soft / cool and cinematic / vivid and bright] color grade without overdoing it.
Keep: Composition, people and objects must not change; only color and tone should change.
Output: A photo with the same size and aspect ratio.

Object removal, background and sky prompts

Remove this from the attached photo: [the person walking by on the right in the background / the trash can / the power line / the plastic bag on the ground].
Fill in whatever should be behind the removed area naturally: the [paving stones / sea / wall texture] should continue without a break.
Light direction, shadows and perspective must match the surroundings; leave no blurry smudge that shows something was erased.
The [person / product] in the foreground and everything else in the photo must stay the same.
Do not add or remove any other object.

The background in the attached portrait is messy and distracting.
Simplify the background without touching the person at all: remove [the people, signs and brightly colored objects behind them].
Then give the background a soft shallow depth of field (bokeh), as if shot with an 85 mm f/1.8 lens; the person must stay sharp.
Hair strands and shoulder edges should look natural, with no hard edge that looks cut out and pasted.
Do not change the light direction or the color temperature.
Output: [vertical 4:5 / original ratio].

The sky in the attached photo is blown out and completely white.
Replace the sky with [an orange-pink sky close to sunset / a light blue sky with scattered clouds / a dramatic overcast sky].
The light of the new sky should fall on the scene too: the color tone of the buildings, water and ground should match the sky, and the shadow direction should stay consistent.
Tree branches and building edges should blend naturally into the sky, with no halos along the edges.
People and objects in the foreground must not change.

Sharpening and colorizing prompts

The attached photo is blurry because it [is slightly shaken / is low resolution / was taken with an old phone].
Sharpen the photo and increase its resolution: sharpen the edges, reduce noise and make the [face / sign / product label] clearer.
Do not redraw the face; the shape of the eyes, nose and mouth and the facial proportions must stay as in the original.
Do not fill in unreadable text with guesses; leave any part that can't be read in the original as it is.
I don't want an overly smooth, plastic-looking texture; keep the natural photo texture.

Colorize the attached black-and-white photo with natural colors that fit [1970s Turkey].
Keep skin tones realistic and different from person to person; pick clothing and object colors from the colors common in that era.
Colors should be light and low in saturation, like analog film; they should not look modern and glossy.
Do not change the faces, expressions or composition; keep the photo's grain.
After giving the image, write in one sentence which colors you chose by guessing.

Portrait prompts: skin retouching and changing clothes

Do a light, natural retouch on the attached portrait.
Fix only these: [pimples and redness / dark circles under the eyes / shine on the forehead].
Keep the skin texture, moles, freckles and facial features; the person should look like themselves, with no filtered look.
Do not slim the face, do not enlarge the eyes, do not lighten the skin tone.
The background, hair and clothing must not change.
Amount: Keep it as restrained as a professional photographer's few-minute retouch.

Change the clothes of the person in the attached photo.
New outfit: [a navy suit and white shirt / a cream knit sweater / a black evening dress].
The fabric should have wrinkles and shadows that fit the body's pose and the light in the scene; the outfit must not look pasted on.
Face, hair, hands, pose, background and lighting must stay exactly the same.
[Keep / Remove] the accessories.

Product and property photo prompts for businesses

The attached photo is a phone shot of a [product name]; prepare it for an e-commerce site.
1. Make the background clean, pure white; leave a light, natural contact shadow under the product.
2. Fix the perspective: the product should face the camera straight on, and vertical lines should be upright.
3. Clean off dust, fingerprints and unwanted reflections on the surface.
4. The product's color, logo, label text, shape and proportions must stay exactly the same; do not show the product as different from what it is.
Output: 1:1 square, with the product filling about 85% of the frame.

Edit the attached [living room / kitchen / office] photo for a listing.
Light: Brighten the room as if it were getting daylight; the windows must not blow out to white, and the outside should stay visible.
Tidy up: Remove [the items on the table, the cables, the bag on the floor]; furniture and walls stay as they are.
Geometry: Straighten the vertical lines so the walls don't look tilted.
Honesty: Do not change the size of the room, the number of windows, or the floor and wall materials; do not add any feature that isn't there.
Output: Horizontal 3:2, in natural colors.

Tool-specific prompts: ChatGPT's selection tool and step-by-step editing in Gemini

Prompt 12 is the short version you type in ChatGPT after clicking the image and marking the area with the selection tool. Since you've already selected the area, two lines are enough. Prompt 13 uses Gemini's ability to remember the previous image in the same chat: you set the rules once in the first message, then ask for one change per message.

In the area I selected, do only this: [make the mug dark green].
Do not touch anything outside the selection; match the light, shadow and texture to the surroundings.

We're going to edit this photo together in a few steps. Rules:
- In each message, make only the change I ask for, and don't undo what you did in earlier steps.
- The person's face, facial features and expression must stay the same as in the attached original photo at every step.
- Do not change the framing or the aspect ratio unless I say so.
Step 1: For now, fix only the exposure and white balance.
(Next message) Step 2: Now remove the [object] in the background and don't change anything else.
(Next message) Step 3: Finally, apply a [warm / cinematic] color tone and deliver it in [4:5] ratio.

The 5 Locks of a Good Photo Editing Prompt

An editing prompt gives consistent results when it answers five questions: what will change, what will stay, how the new element will match the scene, which side effects you don't want, and where the output will be used. All 13 prompts above follow this skeleton, and you can use the table below as a checklist when you write your own.

The most important row in the table is 'Keep'. Image models redraw the area they edit, so anything you don't tell them to keep is open to reinterpretation. In our experience, filling this row with concrete names in photos that include faces, hands, logos or text is the single change that improves results the most.

ChatGPT or Gemini? Which One for Which Edit

Both tools edit an uploaded photo from a written prompt, in Turkish as well as English, but each stands out at different jobs. On the Gemini side, Google has presented keeping a person's likeness across edits as the headline feature since its August 2025 editing update, and Nano Banana Pro, announced in November 2025, added local control over camera angle, focus, color grading and scene lighting. On the ChatGPT side, the selection tool makes it easy to mark a small area and change only that.

Masking with GPT Image is entirely prompt-based. The model uses the mask as guidance, but may not follow its exact shape with complete precision.

Here is the practical advice. If you're changing a narrow area, like the color of a mug, an object on a sign or a single piece of clothing, start with ChatGPT's selection tool. If you're editing a portrait from start to finish through light, background and tone steps, Gemini's multi-turn chat gives steadier results. By OpenAI's own account even a selection isn't a hard boundary, so in both tools you still need to write what to keep into the prompt.

Editing Glossary for Stronger Prompts: 8 Terms

Both tools understand everyday descriptions, but adding the photography term in parentheses makes it easier for the model to know exactly which operation you mean. The rule is simple: say what you want in plain words, then add the name of the technical operation.

The 'upscale' row deserves special attention. Increasing resolution draws plausible detail by guessing, and information that was lost stays lost. So a license plate, a line of text or a facial detail in a sharpened photo may be something that wasn't in the original. AI sharpening should not be used on photos meant as evidence or official documents.

7 Techniques for Better Results

• One prompt, one change: Ask for light, color and cleanup steps in separate messages; that way you also see which step broke the result.
• Order matters: Exposure and white balance first, then object removal and background, and color tone and sharpness last. If tone isn't left for last, the later steps will undo it.
• Make the keep list concrete: Instead of 'keep everything the same', write 'keep the face, hair, clothing, logo and framing the same'.
• Add a reference image: Instead of describing a color tone you like, attach a sample photo and say 'apply the tone and light of this photo'; the Gemini API accepts up to 14 reference images in a single request.
• Say where it will be used: Naming the target, such as an Instagram post, an e-commerce product page or print, gets you the right ratio and framing.
• Zoom in to check: Hands, fingers, glasses and text can look right on a small screen and turn out broken when you zoom in.
• Reset if it breaks: Many rounds on the same image pile up small drifts. If it still isn't fixed after the fifth round, start a new chat with the original.

These seven techniques have one thing in common: they leave the model nothing to guess. In AI photo editing, a bad result usually comes from a gap the prompt left open, and the model's ability is rarely the problem. If you don't fill the gap, the model fills it with its own interpretation.

6 Common Mistakes in AI Photo Editing

1. Vague prompts: Phrases like 'make it look better' or 'make it professional' lead the model to reinterpret the whole photo; write what will change in measurable terms.
2. Compressed files: A photo saved from a messaging app or captured as a screenshot has already lost detail before editing starts.
3. Endless rounds on the same image: Each round adds a small drift; after ten rounds the face can turn into someone else.
4. Expecting miracles from sharpening: Unreadable text or a license plate may look readable after sharpening and still be made up.
5. Showing a product as something it isn't: An edit that changes the color, size or material leads to returns and trust problems.
6. Editing without consent: Uploading and changing someone else's photo without permission carries ethical and legal risk.

The last two mistakes on the list are about responsibility. The more convincing an AI edit gets, the more it matters whether the photo still represents reality. In commercial use especially, that line directly affects customer trust.

Using AI Photo Editing in Turkey: KVKK, Watermarks and Product Photos

Under Turkey's Personal Data Protection Law (KVKK), a person's photo is personal data because it can identify that person, and when a facial image is processed as biometric data, it counts as special category personal data under Article 6 of the Law. That is why getting permission before you upload a photo of a friend, employee or customer to an editing tool is an obligation, especially in business use. This is not legal advice; for company processes, consult your legal counsel.

Google's position on watermarks is clear: images created or edited in the Gemini app get a visible watermark and an invisible SynthID watermark. According to the Nano Banana Pro announcement, the visible watermark stays on the free and Google AI Pro plans and is left off images made by Google AI Ultra subscribers, while SynthID is present on every plan. If you plan to use an image in an ad or in company communications, account for this watermark from the start.

In AI photo editing, the prompt decides the result more than the tool does. When you write what will change in measurable terms and what will stay in concrete terms, both ChatGPT and Gemini fix the photo you have without breaking it. Start with the 13 prompts above, ask for one change at a time, and if someone else's face is involved, get their permission before you upload. If you want to create a brand-new photo or bring two photos together in one frame, our guides to Gemini photo prompts and combining photos cover the other two parts of this cluster.