50+ AI photo editing prompts, grouped by what you actually need

Copy what you need from the five groups below. The formula for writing your own is further down, after the library, along with the three places these prompts should not be pointed.

Editing a photo you have vs. generating a new one

Two different jobs, and picking the wrong one is the most common way to waste a run.

Editing starts with your photo. You upload it and describe a change — swap the background, remove an object, alter a color, restyle it. Everything you didn't mention should survive.

Generating starts with nothing. You describe an image and get a new one. No upload, no original to preserve.

An editing instruction handed to a generator produces something unrelated to your photo, because there was no photo. Each prompt below is tagged [Edit] or [Generate] so you don't have to guess.

A · Headshots and portraits

These run on a photo of someone who agreed to be photographed — usually you. Likeness preservation is best-effort: the model tries to keep a face recognizable, and it does not guarantee it, so check the result against the original before you use it anywhere that matters.

1. [Edit] "Replace the background with a plain light gray studio backdrop, keep the subject's face and hair as they are." → A profile photo out of an ordinary snapshot.
2. [Edit] "Relight this portrait as if from a large softbox at 45 degrees to the left, and soften the hard shadow under the chin." → Flat phone-flash lighting gains shape.
3. [Edit] "Change the shirt to a charcoal crew-neck sweater, keep the pose and the drape natural." → A second look without a second shoot.
4. [Edit] "Add a subtle rim light behind the head so the hair separates from the background." → The subject stops sinking into a dark backdrop.
5. [Edit] "Re-frame to a head-and-shoulders crop with a little space above the head." → Correct framing for a profile picture.
6. [Edit] "Warm the color balance slightly and lift the shadows, leave the background as it is." → Rescues a dim indoor photo.
7. [Edit] "Convert to high-contrast black and white with strong side light, keep skin texture visible." → A speaker-bio portrait with some weight.
8. [Generate] "Editorial headshot of a woman in her forties, natural window light, blurred modern office behind her, neutral palette." → A team-page placeholder while the real shoot is booked.
9. [Generate] "Environmental portrait of a carpenter in a workshop, late afternoon light through a dusty window, warm tones." → An about-page image with a sense of place.
10. [Generate] "Candid group photo of four colleagues laughing in a bright conference room, natural light, wide framing." → A culture-page image.

Retouching skin is a purpose-built tool rather than a prompt — steadier results, and it keeps the texture. That's covered in smoothing skin without the plastic look.

B · Product and marketplace shots

1. [Edit] "Replace the background with pure white and keep a soft contact shadow under the product." → Main-image look without a lightbox.
2. [Edit] "Place this product on a light oak surface with a linen napkin, soft daylight from the left." → A lifestyle secondary image.
3. [Edit] "Remove the price sticker from the box and rebuild the printed surface underneath." → A usable listing photo without a reshoot.
4. [Edit] "Even out the lighting on the product and remove the reflection of the room in the glass." → Glassware that reads clearly.
5. [Edit] "Change the background to a soft pink gradient, keep the product and its shadow unchanged." → A seasonal variant of an existing shot.
6. [Edit] "Straighten the label and smooth the crease in the fabric behind the product." → A tidier catalog frame.
7. [Edit] "Crop to a square with the product centered and filling most of the frame." → The aspect ratio marketplace grids want.
8. [Generate] "Flat lay of desk stationery on a pale gray surface, even overhead light, generous negative space." → A category banner or article header.
9. [Generate] "A ceramic pour-over coffee dripper on a marble counter, morning window light, steam rising, shallow depth of field." → A menu or wholesale-deck image.
10. [Generate] "White running shoes on gray concrete, hard afternoon light, three-quarter angle, streetwear mood." → A social ad image.
11. [Generate] "Overhead shot of a gift box with a satin ribbon on a deep green surface, warm holiday lighting." → A seasonal promotional image.

One caution before any of these reach a listing: marketplaces regulate the main image tightly, and a generated or heavily restyled image is generally not acceptable as the main image of a physical product. The published rules are in product photo tips for online sellers.

C · Travel and landscape

1. [Edit] "Remove the power lines crossing the sky and rebuild the clouds behind them." → The overlook shot you didn't post.
2. [Edit] "Deepen the sunset color slightly and keep the foreground exactly as photographed." → An honest lift rather than a sky swap.
3. [Edit] "Remove the trash can and the parking sign from the left edge of the frame." → A cleaner scenic pull-off.
4. [Edit] "Reduce the haze in the distant hills and recover detail in the sky." → Rescues a flat midday photo.
5. [Edit] "Remove the people walking through the far background, keep the two in the foreground." → A landmark that looks unhurried.
6. [Edit] "Shift this beach photo to a cooler, overcast mood while keeping the composition." → The same frame in a different season.
7. [Edit] "Straighten the horizon and fill the corners the rotation left empty." → Level, without cropping in.
8. [Generate] "Red rock canyon at golden hour, long shadows, hikers small in frame for scale, wide angle." → A trip-recap header.
9. [Generate] "Empty two-lane highway through high desert at dawn, distant mountains, cool blue light." → A road-trip section divider.
10. [Generate] "Small-town main street with brick storefronts and string lights at dusk, warm windows, light rain." → A nostalgic travel-blog image.
11. [Generate] "Pine forest trail after rain, low mist between the trunks, soft even light, tall vertical frame." → A phone wallpaper or story background.

Wire and object removal is steadier with a dedicated tool than with a prompt, and it leaves the rest of the frame alone — see removing power lines from landscape photos.

D · Art and creative

1. [Generate] "Watercolor of a tabby cat asleep on a windowsill, loose washes, soft edges, warm afternoon palette." → A print or greeting card.
2. [Generate] "Flat-design illustration of a small team around a table, pastel palette, simple shapes." → Article artwork.
3. [Generate] "Cyberpunk street at night in the rain, neon in teal and magenta, reflections on wet asphalt." → A poster or desktop background.
4. [Generate] "Minimalist line drawing of a single eucalyptus branch on cream, generous negative space." → A gallery-wall print.
5. [Generate] "Retro 1980s poster with geometric neon shapes, purple and orange gradient, grainy print texture." → An event flyer background.
6. [Generate] "Fantasy illustration of a castle above cloud banks, dramatic backlight, fine detail." → A book-cover concept.
7. [Generate] "Children's book illustration of a boy and a fox crossing a stream, soft gouache, friendly palette." → Story artwork.
8. [Edit] "Convert this photo into a classic oil painting with visible brushwork." → Wall art from a family photo.
9. [Edit] "Turn this photo into a pencil sketch, black and white, visible hatching." → A hand-drawn effect.
10. [Edit] "Restyle this landscape as a mid-century travel poster, flat color blocks, limited palette." → The national-park-poster look.

E · Cleanup and fixes

1. [Edit] "Remove the stranger in the background and fill in the wall behind them." → A frame with only the people who belong in it.
2. [Edit] "Remove the text and logo from the lower right corner and rebuild the background." → A reusable image.
3. [Edit] "Change the cluttered background to a plain cream wall, keep the person as they are." → Attention back on the subject.
4. [Edit] "Remove the cast shadow behind the product so the background reads as one even tone." → A cleaner catalog shot.
5. [Edit] "Smooth the wrinkles in the fabric backdrop, leave the subject untouched." → A studio look out of a bedsheet.
6. [Edit] "Change the jacket from red to off-white, keep the fabric texture and folds." → A colorway preview before you order stock.
7. [Edit] "Change the time of day from midday to blue hour and add warm light in the windows." → A new mood from one frame.
8. [Edit] "Remove the leash from the dog and rebuild the grass underneath." → An uncluttered pet photo.
9. [Edit] "Remove the reflection of the photographer in the mirror." → An interior shot without the giveaway.
10. [Edit] "Remove the recycling bin at the right edge and extend the hedge behind it." → A cleaner exterior — but see the section below before you do this to a property listing photo.

For plain removals — a person, an object, text, a wire — a dedicated removal tool costs less and disturbs the rest of the image less than asking a general editor to redraw the frame.

Write your own: the five-part formula

Every prompt above is built from the same five slots. Fill them and you can generate an unlimited number of your own.

1. Subject — what the image is of.
2. Action or state — what it's doing, or how it sits.
3. Style — editorial, documentary, flat lay, watercolor, cinematic.
4. Light — window light, golden hour, hard afternoon sun, high-key studio, blue hour.
5. Composition and detail — angle, background, aspect ratio, palette, depth of field.

Concrete nouns and adjectives beat evaluative ones. "Make it better" gives the model nothing to act on; "soft daylight from the left, plain background, warm palette" gives it three decisions.

Style vocabulary is worth collecting, because a single term can carry a whole look: golden hour, blue hour, flat lay, bokeh, high-key, moody, three-quarter angle, negative space, shallow depth of field.

Change one part at a time

When a result is close but wrong, don't rewrite the prompt. Change one slot — just the light, or just the style — and run it again. Rewriting from scratch changes several variables at once, so you learn nothing about which one was responsible. There's randomness in the output too, which means an identical prompt can give a different result on a second run.

Before you start

Four things worth knowing so nothing surprises you.

There's a text box, and it has a limit. Prompts go up to 1,000 characters. That's generous for the five-slot formula and tight if you paste an essay. Empty prompts are rejected.

Safety checks run before anything is charged. Prompts are screened, and one that's refused is stopped before any credit moves and before a job is created. A refusal costs you nothing but the time.

Results vary. Same prompt, same image, different run — expect some variation, especially on faces and hands. Check at full size, not at thumbnail size.

Purpose-built tools beat prompts for narrow jobs. Background removal, object removal and skin retouching each have a dedicated tool. They're steadier and they leave the rest of the photo alone. Prompts earn their keep when the change is descriptive rather than mechanical.

Where not to use these prompts

Three hard boundaries, all specific to the United States, and all worth knowing before rather than after.

Passport and visa photos. Not one line in this library belongs near a document photo. The State Department wants the frame as the camera made it, treats software, phone apps, filters and AI as one category of change you may not make, and says submitted photos are screened for signs of AI tools. A prompt is that category by definition — which puts the whole library out of scope here, including the mild-sounding relighting lines in group A. Even the gentlest cleanup counts: the guidance rules out fixing red eye afterwards. Reshoot instead. The document-photo boundary is set out in full in the skin-retouching article linked from group A above.

Real estate listing photos. Prompt 52 is the exact edit MLS rules were written to stop. The California Regional MLS tells subscribers that a listing image may not be altered to take elements out of the property, and overhead poles and wires are named among the elements it means — so the prompt that would most improve a listing photo is the one that can pull the listing and land the agent a complaint. California has also required, since 01/01/2026 under AB 723, that a significantly altered listing image be disclosed and shown beside its unaltered original; disclosing an edit is not the same as being allowed to make it. If listing photos are your work, read the wire-removal article linked from group E before you use any prompt in this file on one.

Marketplace main images. Amazon's published apparel guide bars logos, text, graphics and badging that aren't part of the product from the main image, and requires a pure white background with the product filling at least 85% of the frame. Restyled or generated images don't belong in that slot.

There's a fourth boundary that isn't about a rulebook. Don't use these prompts to put a real, identifiable person into an image they never agreed to — and nothing here should be pointed at making suggestive imagery of a recognizable person. That's why this library has no face-transplant prompt in it. Editing your own photo, or a photo taken with everyone's agreement, is the assumption throughout.

Who owns an AI-edited image in the US?

Relevant if you plan to sell what you make.

The US Copyright Office's Part 2: Copyrightability report, published 01/29/2025, concluded that "the mere provision of prompts" does not qualify for copyright protection. Protection is available where "a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output" — and using AI in the process, or including AI-generated material in a larger human-made work, does not by itself bar copyrightability. The Office also found the case had not been made for changing existing law.

Practical reading: editing your own photograph sits differently from typing a description and receiving a new image, because in the first case your photograph is perceptible in the output. This is a pointer to the report, not legal advice, and it's assessed case by case.

Do you have to label AI-edited photos?

It depends on where the image goes.

On YouTube, creators must disclose altered or synthetic content that is realistic — "content a viewer could easily mistake for a real person, place, scene, or event." Disclosure isn't required for clearly unrealistic content such as animation, for color adjustment, lighting filters and special effects, or for using AI in production tasks like scripts, ideas and captions. The label sits in the expanded description, and moves onto the video itself for sensitive topics.

In California, the AI Transparency Act (SB 942, amended by AB 853, which pushed the effective date to 08/02/2026) requires covered providers — generative AI systems with over 1,000,000 monthly visitors or users that are publicly accessible in California — to offer a free detection tool and to give users the option of a visible disclosure and a hidden provenance signal. Related duties reach large online platforms from 01/01/2027 and capture-device manufacturers from 01/01/2028.

The distinction that matters: those obligations sit on providers, platforms and manufacturers, not on you as an end user. California is not requiring you to label your own photos.

Frequently asked questions

Can I copyright an AI-edited image?

Per the Copyright Office's January 2025 report, prompts alone don't qualify. Protection can attach where a human-authored work is perceptible in the output, or where a person makes creative arrangements or modifications of it, assessed case by case. Dedicated pointer, not legal advice.

Do I have to label AI-edited photos?

It depends on the surface. YouTube requires disclosure for realistic altered or synthetic content. California's SB 942, as amended by AB 853 and effective 08/02/2026, places its duties on large providers rather than on end users.

Why isn't my prompt working?

Usually a missing slot. Check all five — subject, action, style, light, composition — then change exactly one and rerun.

Can I use these on a passport or ID photo?

No. The State Department prohibits software, app, filter and AI changes to passport photos and checks for them.

What's the difference between editing and generating?

Editing changes a photo you upload and keeps everything you didn't mention. Generating makes a new image from a description, with no original involved.

Pick one and run it

Take the prompt closest to what you need, change one slot to match your photo, and run it once. That's a faster way to learn what the model responds to than reading another list.

How this article was made

• Written — seo-content produced the en-US version of /blog/prompt-chinh-sua-anh-bang-ai under brief I18N-11, 2026-08-29.
• Reviewed — seo-review re-ran the §G publication gate on 2026-09-06 (VIPIX session s148) on an earlier revision and returned PASS-WITH-FIXES, and the next revision fixed the issues that verdict names. Verdict on file: reviews/en-US-regate-s148-verdict.md (earlier gates, 2026-08-29: en-US-T1-regate-verdict.md, en-US-T2T4-gate-verdict.md, en-US-T3-gate-verdict.md). Since then, the article's internal links have been updated, and on 2026-09-28 this note was corrected.
• Process — content drafted by the Vipix SEO team's LLM under a reviewed brief; edited and checked by the Vipix SEO team; no native-speaker contractor was engaged.