You ask ChatGPT to remove one person standing behind you.
It removes them.
Great.
Then you look at your own face and think:
Wait… who is that?
That is the frustrating part of AI photo editing. Getting ChatGPT to understand what should change is only half the job.
You also need to tell it what it has no permission to change.
ChatGPT Images can edit uploaded photos, remove or add elements, make backgrounds transparent, work on selected areas and regenerate images in different aspect ratios. OpenAI also recommends making small, specific revisions and clearly stating what should remain unchanged. OpenAI image editing guidance
That is the approach I’m using throughout these 30 prompts.
Not:
“Make this photo better.”
Instead:
“Change this. Leave these details alone. Make the new area match the original. Give me this final format.”
That small change in thinking can save you a lot of unnecessary reruns.
What Do You Want ChatGPT to Fix?
| Your goal | Start with |
|---|---|
| Clean up a photo without changing the subject | Prompts 1–5 |
| Improve a portrait or headshot | Prompts 6–10 |
| Edit product and ecommerce photos | Prompts 11–15 |
| Change backgrounds, rooms or composition | Prompts 16–20 |
| Repair or restore older photos | Prompts 21–24 |
| Reuse images for social media and design | Prompts 25–30 |
Can ChatGPT Really Edit Photos?
Yes. You can upload an existing image and describe the exact change you want, or select part of the image before giving your instruction. ChatGPT can remove, add and update image elements, create transparent backgrounds and regenerate images in a different aspect ratio.
The current editor gives you two basic choices: describe the edit directly or highlight a specific area first. OpenAI does warn that a highlighted area is not always exact and an edit can spread beyond your selection. OpenAI ChatGPT Images help
That is why I would not rely on selection alone.
If I circle a lamp and want it removed, I would still write:
Remove only the selected lamp. Keep the wall, floor, furniture, lighting, perspective and every other object unchanged.
Think of the selection as showing ChatGPT where to look.
Your prompt still tells it what it may touch.
Why Does ChatGPT Change My Face When I Edit a Photo?
Generative photo editing can recreate parts of an image rather than working like an old-fashioned pixel editor. Even when you request one small change, nearby details can sometimes shift as ChatGPT produces the new version.
OpenAI says its current image model is better at preserving likeness, composition and lighting during edits, but “better” does not mean every edit will remain identical. OpenAI image model update
If the photograph contains your face, a product label, artwork, a logo or another detail that matters, name it.
Do not assume ChatGPT understands that your face is sacred but the ugly lamp behind you can go.
How Do I Stop ChatGPT From Changing My Face?
Tell ChatGPT exactly which identity details must stay unchanged, then keep the edit itself narrow. The more the face matters, the less freedom I would give the model around it.
A useful identity lock is:
IDENTITY LOCK: Keep the exact same facial structure, skin tone, natural skin texture, expression, eye shape, nose, lips, jawline, hairstyle, hairline, body proportions and pose. Do not beautify, reshape, age, de-age or reinterpret the person.
Then describe the change.
For example:
Change only the background to a softly lit modern office. Keep the person exactly as described in the identity lock.
That wording closely follows OpenAI’s own current advice to explicitly state what should change and what should stay the same. OpenAI image prompting guide
What Is the Best ChatGPT Photo Editing Prompt Formula?
I use five pieces: what is locked, what should change, where that change belongs, what the new edit must visually match and what final format I want. This gives ChatGPT boundaries without turning the prompt into a novel.
The 5-Part Photo Edit Formula
| Part | Tell ChatGPT |
|---|---|
| LOCK | What cannot change |
| EDIT | What should change |
| AREA | Where the change belongs |
| MATCH | Lighting, shadows, scale, perspective, texture |
| OUTPUT | Crop, aspect ratio, transparency or final use |
Example
LOCK: Keep my face, hair, expression, pose and clothing exactly unchanged.
EDIT: Remove the two people standing behind me.
AREA: Background only.
MATCH: Rebuild the missing wall and pavement using the same perspective, texture, lighting and shadows already present.
OUTPUT: Keep the original vertical composition.
You can copy that structure for almost every prompt below.
Which ChatGPT Prompts Can Fix a Photo Without Changing the Subject?
These first prompts are for photos that are already usable except for one annoying problem. The goal is correction, not reinvention.
1. How Do I Remove a Background Without Ruining Hair?
Use this when you need a clean cutout for a profile image, product, Canva design or website. Fine hair and soft edges need to remain part of the subject rather than being chopped away with the background.
Copy-Ready Prompt
Remove only the background from this image and replace it with full transparency. Keep the main subject exactly unchanged, including the face, hair, flyaway hairs, clothing, body shape, hands, edges, colour and original detail. Preserve fine hair strands and soft natural edges rather than cutting them into a hard outline. Do not add a new background, shadow or object. Final result: clean transparent background suitable for reuse in another design.
Why this wording helps
“Remove the background” tells ChatGPT what disappears.
The rest tells it what needs to survive.
If it goes wrong
Return to the original image and try again rather than repairing a cutout where the hair has already been changed.
2. How Do I Remove a Person From the Background?
This works for travel photos, events and street shots where an unwanted person is distracting from the main subject. The reconstruction should happen behind you, not through you.
Copy-Ready Prompt
Remove only the person standing [DESCRIBE LOCATION, e.g. “behind my left shoulder”]. Keep the main subject completely unchanged, including face, expression, hair, clothing, pose, hands and body. Reconstruct only the background area that becomes empty. Match the existing wall, floor, objects, lighting, shadows, perspective, depth and image grain so the removal is not obvious. Do not crop, zoom or reposition the main subject.
Social-media use
This can rescue an otherwise good vacation or event photograph without forcing you to crop away half the scene.
If it changes you
Use the selection tool on the unwanted person, then repeat the preservation instructions. OpenAI notes that selections can still affect nearby areas, so keep checking the main subject afterward. OpenAI image editor instructions
3. How Do I Fix Harsh Lighting Without Changing My Face?
Ask for relighting rather than retouching. The job is to soften the light falling across the existing person, not create a smoother, younger or more symmetrical version of them.
Copy-Ready Prompt
Correct only the harsh lighting in this portrait. Keep the person’s facial structure, natural skin texture, skin tone, expression, hair, clothing, pose and body exactly unchanged. Soften the strongest facial shadows and overly bright highlights while preserving realistic depth and natural contrast. Keep the original direction of light believable. Do not smooth skin, reshape the face, change eye colour, whiten teeth or beautify the person.
Why this matters
“Improve the face” gives the model broad permission.
“Correct only the harsh lighting” gives it a job.
4. How Do I Brighten a Dark Photo Without Making It Look Fake?
The aim is to reveal detail that is already plausible in the photograph without turning a night scene into midday. Keep the mood and colour relationships intact while making the subject easier to see.
Copy-Ready Prompt
Brighten this dark photo conservatively while keeping the original scene, mood and subject unchanged. Lift shadow detail enough to make the person and surroundings easier to see, reduce excessive darkness and balance the exposure. Preserve realistic blacks, natural skin tone and the existing light direction. Do not make the scene look like daylight, flatten all shadows, change the person’s face or add new light sources.
Useful for
Indoor phone photos, restaurants, evening events and rooms where the camera exposed for the wrong part of the scene.
5. How Do I Make a Blurry Photo Look Clearer?
ChatGPT can make a slightly soft photograph look cleaner, but it cannot truly recover every detail the camera failed to record. Faces, tiny text and product labels deserve extra checking because reconstructed detail can look convincing while still being wrong.
Copy-Ready Prompt
Improve the clarity of this slightly blurry image conservatively. Reduce visible softness and noise, improve edge definition and make existing details easier to see. Keep the exact face, expression, product details, text, colours, proportions and composition unchanged. Do not invent facial features, lettering, patterns or object details that are not clearly supported by the original image. Keep the result natural rather than oversharpened.
When I would stop
If the original face is only a handful of blurry pixels, I would not treat the generated result as recovered truth.
It is an interpretation.
What Are the Best ChatGPT Prompts for Portraits and Headshots?
For portraits, I would lock the person first and edit around them. Background, crop, wardrobe and light can change; identity should not casually join the makeover.
6. How Do I Turn a Selfie Into a Professional Headshot?
A useful headshot should still look like the person someone will meet. I would improve the environment, framing and light before asking ChatGPT to “improve” the face.
Copy-Ready Prompt
Turn this selfie into a clean professional headshot while preserving my exact identity. Keep my facial structure, age, skin tone, natural skin texture, eye shape, nose, lips, jawline, hairstyle and body proportions unchanged. Improve the framing to a professional chest-up portrait, use balanced flattering light and replace the existing background with a simple [neutral studio / softly blurred office] background. Keep the clothing unchanged. Do not beautify, slim, reshape, de-age or alter my expression.
Best for
LinkedIn, company bios, speaker profiles and professional websites.
Before using it
Compare it side by side with the original.
If a recruiter could meet you and wonder where your AI-generated twin went, the edit went too far.
7. How Do I Change a Headshot Background Without Changing My Face?
This is one of the safer portrait edits because the requested change can stay behind the person. Still, spell out that hair, facial features and edge details belong to the locked subject.
Copy-Ready Prompt
Replace only the background of this headshot with [DESCRIBE BACKGROUND]. Keep the person completely unchanged: same face, expression, skin texture, hair, hairline, clothing, pose and body proportions. Match the new background lighting to the existing light on the person. Create believable depth and natural separation around hair and shoulders. Do not recolour or relight the face unless needed only to make the background integration believable.
Background ideas
A soft grey studio.
A bright office.
A warm cream backdrop.
A subtle outdoor background.
One change.
Four different uses.
8. How Do I Change Clothes in a Photo Without Changing the Person?
Lock everything about the person except the garment you want replaced. Clothing sits close to hands, hair and body shape, so I would describe those boundaries carefully.
Copy-Ready Prompt
Change only the [SHIRT/JACKET/DRESS] into [DESCRIBE NEW GARMENT]. Keep the exact same person, face, expression, skin tone, hairstyle, body shape, pose, hands and background. Make the new garment fit the existing pose naturally with realistic fabric folds, seams, shadows and contact with the body. Do not change body proportions or exposed skin. Keep everything outside the clothing area unchanged.
Useful for
Visualising a professional outfit, colour option or styling direction.
Remember
It is a visual preview.
A generated jacket does not know what the real jacket will feel like, fit like or look like under your actual lighting.
9. How Do I Preview a Different Hair Colour Without Changing My Face?
Tell ChatGPT to preserve the haircut, texture and length and change only the colour. This keeps the experiment focused on the decision you are actually trying to preview.
Copy-Ready Prompt
Change only my hair colour from [CURRENT COLOUR] to [NEW COLOUR]. Keep the exact haircut, hair length, curl or wave pattern, density, hairline and styling unchanged. Keep my face, eyebrows, skin tone, expression, clothing, pose, lighting and background exactly the same. Make the new hair colour respond naturally to the existing highlights and shadows. Do not change facial features or makeup.
Try this
Generate two or three colour choices from the same original.
Do not edit Version A into Version B into Version C.
Return to the source each time so you are comparing hair colours rather than three generations of facial drift.
10. How Do I Retouch Skin Without Getting Plastic AI Skin?
Ask for temporary distractions to be softened while natural texture stays visible. If the prompt tells ChatGPT to make the skin flawless, you may get exactly that—and lose the person in the process.
Copy-Ready Prompt
Retouch the skin lightly while preserving realistic skin texture, pores, natural lines, freckles and the person’s exact facial structure. Reduce only [TEMPORARY BLEMISHES / MINOR REDNESS / UNEVEN PATCHES] without changing skin tone or making the face look airbrushed. Keep the eyes, nose, lips, jawline, expression, hair and lighting unchanged. Do not reshape the face, whiten teeth, enlarge eyes or create plastic-looking skin.
My rule
Temporary blemish?
Reasonable edit.
Entire face becoming a polished 3D character?
Start again.
What Are the Best ChatGPT Photo Editing Prompts for Product Images?
For ecommerce work, the product should usually be the locked object while the scene around it changes. If buyers need to know what they will actually receive, labels, colours, shape, size and materials should remain faithful to the source.
11. How Do I Put a Product on a Clean White Background?
This creates a cleaner listing image while leaving the product itself alone. Ask for a realistic contact shadow so the object does not look as though it is hovering two centimetres above the table.
Copy-Ready Prompt
Keep this product exactly unchanged, including its shape, proportions, colour, material, label, logo, packaging, cap, text and visible details. Remove the existing background only and replace it with a clean pure-white ecommerce background. Keep the product centred and correctly scaled. Add only a subtle realistic contact shadow directly beneath the product so it appears grounded. Do not redesign, simplify or reinterpret any part of the product.
Before publishing
Zoom in on:
logo
label text
colour
cap
edges
small accessories
If those are wrong, the photo is not ready for a product page.
12. How Do I Put a Product Into a Realistic Lifestyle Scene?
A lifestyle image helps the customer picture the item somewhere it belongs. The trick is making the environment adapt to the product instead of allowing the product to be redesigned to fit the scene.
Copy-Ready Prompt
Use the uploaded product as the exact locked reference. Preserve its shape, proportions, packaging, label, logo, colour, material and all visible details. Place it naturally in [DESCRIBE REALISTIC SCENE]. Match the product’s scale to nearby objects and create believable lighting, reflections, contact shadows and depth based on the environment. Do not change the product to match the room. Leave enough breathing room around it for a clean commercial composition.
Examples
Skincare on a bathroom counter.
Candle on a bedside table.
Tote bag beside a café chair.
Notebook on a working desk.
The setting should answer:
Where does this product belong in my customer’s life?
13. How Do I Remove Bad Reflections From a Product Photo?
Remove only the reflection that hides useful product information. Gloss, glass and metal still need some reflections or they may stop looking like the material they actually are.
Copy-Ready Prompt
Remove or reduce only the distracting glare covering [DESCRIBE AREA] on this product. Keep the exact label, logo, lettering, product shape, colour, material, reflections elsewhere and background unchanged. Reconstruct the hidden area only where the original product detail is clearly supported. Preserve enough natural reflection to keep the surface looking like real [GLASS / METAL / GLOSSY PLASTIC]. Do not redesign the packaging or invent unreadable text.
Best practice
If glare completely covers important lettering, use another product reference if you have one.
Do not ask the model to guess your packaging copy.
14. How Do I Create a Product Ad Image With Space for Text?
Composition matters before the headline is added. Place the product where it has room to breathe and deliberately leave one clean area for your actual marketing copy later.
Copy-Ready Prompt
Keep the uploaded product exactly unchanged. Create a [4:5 / 1:1 / 9:16 / 2:3] commercial composition using [DESCRIBE BACKGROUND OR SETTING]. Position the product in the [LEFT/RIGHT/LOWER] portion of the frame and leave a large uncluttered area on the [OPPOSITE AREA] for text I will add later. Match realistic scale, lighting, shadows and reflections. Do not add generated headlines, slogans, logos or extra products.
Why I prefer this
ChatGPT can generate text in images, but for brand campaigns I often want complete control over:
font
spelling
line breaks
brand colour
placement
So I would create the visual first.
Then add the final copy in the design tool.
15. How Do I Create Seasonal Product Photos Without Another Photoshoot?
Keep the product fixed and change the environment around it. This can turn one good packshot into seasonal campaign concepts without covering the item in so many decorations that nobody remembers what you sell.
Copy-Ready Prompt
Use this product as the exact locked reference. Keep its packaging, colour, label, logo, material, size and proportions unchanged. Create a tasteful [CHRISTMAS / AUTUMN / VALENTINE’S / SPRING / MOTHER’S DAY] setting around the product using [DESCRIBE 2–4 APPROPRIATE DETAILS]. Match realistic scale, light direction, shadows and reflections. Keep the product fully visible and make it remain the main subject. Do not cover the label or add text.
Repurpose idea
Use the same clean source for each season.
Do not build Christmas from the Valentine’s edit and then build spring from Christmas.
By spring, your original bottle may have gone through more changes than a soap opera cast.
How Can I Change a Photo Background Without Making It Look Pasted On?
A believable replacement has to agree with the subject’s perspective, light, scale and depth. Changing the location is easy; making the person or object feel as though they were actually photographed there is the harder part.
16. How Do I Replace a Photo Background Realistically?
Tell ChatGPT to rebuild the environment around the locked subject and match visual cues from the original image. Hair edges, feet and contact shadows deserve special attention.
Copy-Ready Prompt
Replace only the existing background with [DESCRIBE NEW LOCATION]. Keep the main subject completely unchanged. Match the new environment to the subject’s existing camera angle, perspective, height, scale, light direction, colour temperature and depth. Create realistic contact shadows and natural edge blending around hair, clothing and feet. Do not move, resize, reshape or relight the subject beyond what is needed for believable integration.
Check these three areas
Hair.
Feet.
Shadow.
They often reveal whether the subject feels pasted into the scene.
17. How Do I Blur a Background Like a Real Camera Lens?
A believable shallow-depth effect keeps the subject sharp while blur increases naturally with distance. The background should not look like someone selected it with a giant blur brush.
Copy-Ready Prompt
Keep the main subject completely sharp and unchanged. Apply realistic shallow depth of field to the background as if photographed with a portrait lens at a wide aperture. Objects immediately behind the subject should be slightly soft, with blur increasing naturally farther into the distance. Preserve believable depth, light, colour and subject edges. Do not blur hair, clothing edges or anything on the same focal plane as the subject.
Use this when
The background is distracting but you still want the viewer to understand the location.
18. How Do I Extend a Photo Into a Different Aspect Ratio?
Extending the scene can protect the subject from an awkward crop. Tell ChatGPT to leave the original image area untouched and generate only the new space needed around it.
Copy-Ready Prompt
Convert this image to [DESIRED ASPECT RATIO] by extending the canvas rather than cropping or enlarging the main subject. Keep every pixel-equivalent detail in the existing image area visually unchanged, including the person, objects, architecture, lighting and composition. Generate only the missing space around the original scene. Continue walls, floor, sky, patterns, shadows and perspective naturally. Do not add new people, text or unrelated objects.
Great for
Turning horizontal photos into vertical posts.
Creating Story versions.
Making banner compositions.
Giving an image more room for text.
19. How Do I Preview a New Wall Colour or Furniture in My Room?
Use the photo as a fixed room reference and change only the design decision you are testing. The result can help you think through an option, but it is still a visual preview rather than a guarantee of exact real-world colour or scale.
Wall-Colour Prompt
Keep this room completely unchanged, including the furniture, flooring, windows, décor, room dimensions, camera position and lighting. Change only the [SPECIFIC WALL] from its current colour to [NEW COLOUR]. Preserve the wall texture and allow the new colour to react naturally to existing highlights and shadows. Do not recolour nearby furniture, ceiling, trim or flooring.
Furniture Prompt
Keep the room, architecture, flooring, windows, existing lighting and camera angle unchanged. Replace only [CURRENT FURNITURE ITEM] with [DESCRIBE NEW ITEM]. Make the new furniture believable for the available floor space, perspective and surrounding objects. Match realistic contact shadows and scale. Do not rearrange the rest of the room.
Smart use
Generate three options from the same original room photo.
Now you have an A/B/C decision instead of trying to imagine paint chips in your head.
20. How Do I Combine Two Photos Into One Realistic Image?
Tell ChatGPT which source controls which element and where each belongs. The more specific the relationship between the two images, the less room there is for faces, sizes and positions to wander.
Copy-Ready Prompt
Combine these two uploaded images into one realistic photograph. Use Image 1 as the exact reference for [PERSON/OBJECT A] and Image 2 as the exact reference for [PERSON/OBJECT B]. Keep each person’s facial identity, hair, clothing and body proportions unchanged. Place [A] on the [LEFT/RIGHT] and [B] on the [LEFT/RIGHT] in [DESCRIBE SHARED SCENE]. Match camera height, perspective, scale, focus, colour temperature, light direction and shadows so both subjects appear photographed together. Do not blend their facial features or alter their identities.
OpenAI’s current prompt guidance recommends explicitly identifying uploaded references by order and explaining how they relate spatially. OpenAI multiple image guidance
What Are the Best ChatGPT Prompts for Restoring Old Photos?
Old-photo repair needs more restraint than ordinary editing. Removing scratches or balancing fading is different from asking AI to invent facial features or historical details that no longer exist in the source.
21. How Do I Remove Scratches and Dust From an Old Photo?
Keep the people and scene fixed and repair only visible surface damage. This is closer to cleanup than reconstruction.
Copy-Ready Prompt
Restore this old photograph conservatively. Remove visible dust, small scratches, spots and surface marks while keeping the original people, facial features, expressions, clothing, objects, background, framing and photographic character unchanged. Reconstruct only tiny damaged areas where surrounding detail makes the repair clear. Do not modernize the image, sharpen faces aggressively or invent features that are missing.
Good candidate
Dust across a plain wall.
Small scratch through clothing.
Minor spotting.
More risky
Large missing section across a face.
That requires guessing.
22. How Do I Restore a Faded Family Photo?
The goal should be making the original easier to see, not making a decades-old photograph look as though it was shot yesterday on a new phone. Preserve the age and character of the image.
Copy-Ready Prompt
Restore this faded family photograph while preserving its original people, faces, expressions, clothing, setting and period appearance. Improve faded contrast and tonal separation gently, reduce colour cast where appropriate and make existing details easier to see. Keep natural film grain and photographic texture. Do not modernize clothing, skin, hair, background objects or facial features. Do not invent detail in areas that are too damaged to support it.
My preference
A restored family photo should still feel like that photograph.
Not a polished remake starring people who merely resemble your relatives.
23. How Do I Colourize a Black-and-White Photo Carefully?
AI can create plausible colours, but unless you have real references, those colours are interpretations. The prompt should favour restrained, period-appropriate choices rather than pretending the model knows the exact original shirt colour.
Copy-Ready Prompt
Colourize this black-and-white photograph conservatively while keeping every face, expression, object, clothing shape, composition and historical detail unchanged. Use restrained, natural colours appropriate to the apparent time period and environment. Preserve original lighting, contrast and photographic texture. Where the true colour cannot be known from the image, choose a plausible muted interpretation rather than an unusually bright or modern colour. Do not alter facial features or add objects.
Better still
If you know:
the dress was blue,
the car was green,
the wall was cream,
tell ChatGPT.
Evidence beats guessing.
24. How Do I Repair a Torn Photo Without Inventing a New Face?
Rebuild missing backgrounds where surrounding information makes the continuation clear. If a large part of a face is missing, treat the generated reconstruction as interpretation rather than a factual restoration.
Copy-Ready Prompt
Repair the torn and missing portions of this photograph conservatively. Preserve every surviving facial feature, expression, hairstyle, piece of clothing, object and background detail exactly as shown. Reconstruct missing background, wall, floor, sky or clothing only where surrounding information clearly supports the continuation. If facial detail is missing and cannot be inferred reliably, do not redesign the person’s face or invent unsupported features. Keep the restored area visually consistent with the original photograph’s grain, contrast and age.
Why this matters
A believable face is not automatically the real face.
For family-history work, that difference matters.
What ChatGPT Photo Editing Prompts Are Best for Social Media?
Decide where the image is going before asking for the edit. A Pinterest Pin, Instagram portrait, Story and YouTube thumbnail need different space around the subject even when they all begin with the same photograph.
25. How Do I Turn a Photo Into a 4:5 Instagram Post?
Ask ChatGPT to create the vertical composition around the existing image rather than automatically chopping off whatever does not fit. Preserve the subject and decide where you want breathing room before adding text.
Copy-Ready Prompt
Recompose this photo into a 4:5 vertical social-media image. Keep the main subject, face, body, clothing and important original objects unchanged. Extend or reposition only the surrounding scene where needed so the subject fits comfortably without cutting off the head, hands or important details. Preserve the existing visual style, perspective and lighting. Leave [TOP/LEFT/RIGHT] area relatively uncluttered for optional text I will add later. Do not generate text.
26. How Do I Turn a Photo Into a 2:3 Pinterest Pin?
Pinterest currently recommends a 2:3 image ratio, commonly 1000 × 1500 pixels. Build the composition for that tall format rather than simply stretching or cropping the original. Pinterest current Pin guidance
Copy-Ready Prompt
Recompose this image as a 2:3 vertical Pinterest image. Keep the main subject and all important original details unchanged. Extend the environment naturally where needed rather than stretching the subject. Position the subject so the composition remains easy to understand at mobile size. Leave a clean area in the [TOP/MIDDLE/BOTTOM] for headline text I will add afterward. Preserve believable lighting, scale, perspective and background continuation. Do not generate text.
Use case
Blog Pin.
Product Pin.
Tutorial cover.
Quote visual.
Before-and-after.
One original photograph can become a different Pinterest asset without pretending it was another photoshoot.
27. How Do I Create a YouTube Thumbnail Background From a Photo?
The visual job is to separate the subject, simplify competing background detail and leave a deliberate area for the title. I would add the final headline myself so typography stays under control.
Copy-Ready Prompt
Turn this photo into a clean 16:9 YouTube thumbnail base. Keep the person’s exact face, expression, hairstyle, body proportions and clothing unchanged. Increase visual separation between the subject and background without creating an artificial cutout. Simplify distracting background objects and create a clean area on the [LEFT/RIGHT] for a headline I will add later. Preserve natural skin and realistic lighting. Do not generate words, logos or clickbait symbols.
One warning
Do not ask for “more shocked expression” unless you actually want ChatGPT changing your expression.
That is not background editing anymore.
28. How Do I Create a Before-and-After Comparison Image?
Keep both sides equally readable so the viewer can judge the actual change. If one side has terrible lighting and the other looks professionally photographed, the comparison stops being useful.
Copy-Ready Prompt
Create a clean side-by-side before-and-after comparison using these two uploaded images. Place the BEFORE image on the left and AFTER image on the right. Preserve both source images accurately without changing faces, products, colours or results. Match their displayed size and alignment so the comparison is fair. Use a simple neutral dividing line between them. Leave a small area for BEFORE and AFTER labels that I will add later. Do not improve or worsen either source.
Useful for
Room edits.
Product changes.
Restoration.
Design experiments.
Photo corrections.
29. How Do I Create a Transparent PNG Cutout?
A reusable cutout should protect fine edges around hair, clothing, products and soft details. The output is useful for design work because you can place the subject into other layouts later.
Copy-Ready Prompt
Create a clean transparent-background cutout from this image. Keep the complete subject exactly unchanged, including facial features, hair strands, clothing, hands, product edges, colours and small visible details. Remove everything outside the subject and make it fully transparent. Preserve natural semi-transparent edge details where appropriate instead of creating a hard artificial outline. Do not add a shadow, coloured background, border or new object.
Use it for
Canva layouts.
Website hero graphics.
Product promos.
Presentation covers.
Social graphics.
One good cutout can save you from repeating the same edit every time.
30. How Do I Turn One Phone Photo Into Several Brand Images?
Use the clean original as the locked reference for every new variation rather than repeatedly editing the latest AI version. This reduces the chance that small differences accumulate across a campaign.
Master Prompt
Use this uploaded image as the permanent locked source reference. Keep the main person/product exactly consistent in every version, including identity, facial features or product shape, colours, logo, materials, proportions and key details. Create separate compositions from this same original for the following uses:
Version 1: 4:5 social post with clean space for a short headline.
Version 2: 2:3 Pinterest composition with vertical background extension.
Version 3: 9:16 Story composition with the subject kept clear of interface areas.
Version 4: wide website banner with the subject positioned on one side and clean copy space on the other.
Version 5: clean standalone product/profile image.
Keep the source subject consistent. Adapt only the surrounding composition, crop and background needed for each format. Do not generate any text.
The real lesson
One original does not have to mean one post.
But every variation should keep looking like it came from the same person, product or campaign.
What Should I Do When ChatGPT Changes the Wrong Part of My Photo?
Do not keep repairing a version that has already drifted far from the source. If the wrong change affects identity, packaging, text, structure or another detail that matters, I would return to the original and try again with tighter boundaries.
OpenAI’s current guidance recommends small revisions one element at a time because targeted changes are easier to follow and help maintain consistency. OpenAI image prompting guide
A useful correction prompt is:
Undo the unintended changes. Return to the original uploaded image as the source. Change only [EXACT EDIT]. Keep [LIST LOCKED DETAILS] exactly the same. Do not crop, zoom, reposition or reinterpret anything else.
If that still drifts?
Start again from the clean source.
Sometimes the best repair is refusing to repair the repair.
Should I Make One Photo Edit at a Time?
Yes when exact details matter. One focused request makes it easier to see what changed and catch a problem before it becomes part of the next version.
Imagine you want to:
remove a person,
change your jacket,
replace the background,
brighten the face.
I would not ask for all four first.
Start with the removal.
Compare.
Then jacket.
Compare.
Then background.
Compare.
Then lighting.
This is slower than one giant prompt.
It is also much faster than reaching Version 7 and realizing your nose quietly changed back at Version 2.
When Should I Use ChatGPT’s Selection Tool?
Use selection when the requested edit belongs to one specific region and the rest of the image should remain stable. Pair the selection with a written boundary because OpenAI says selected edits can extend beyond the exact highlighted area. OpenAI selection tool guidance
Good uses include:
removing one lamp,
changing a shirt colour,
repairing one scratch,
removing one background person,
changing one wall.
Then write:
Change only the selected area. Keep everything outside the selected area unchanged.
Selection narrows the neighbourhood.
Your words still set the rules.
When Should I Restart From the Original Photo?
Restart when something meaningful has already been changed incorrectly. The next edit should not inherit a wrong face, broken product label or altered room layout as its new truth.
My rule is simple:
If the unwanted change is tiny and local, correct it.
If it affects identity, product accuracy or major structure, go back.
Examples worth restarting:
Your face changed.
Logo changed.
Product colour changed.
Furniture moved.
Hands became distorted.
Packaging text changed.
The room was cropped.
You would rather lose one generation than build five more on top of the wrong source.
When Should I Not Trust an AI Photo Edit Blindly?
Be careful when the image is supposed to document what something actually looked like. A generated edit can look believable while including details the original never contained.
I would inspect closely when working with:
family-history photographs
products being sold
logos and branded packaging
documentary images
property-condition photographs
images being used as factual evidence
ChatGPT is useful for creative editing.
“Looks real” and “is historically or commercially exact” are different standards.
That distinction matters most when the image is carrying information rather than simply looking good.
Which ChatGPT Photo Editing Prompt Fits My Goal?
| If you need… | Start with |
|---|---|
| Remove distractions | 1–5 |
| Better portrait/headshot | 6–10 |
| Product listing or campaign image | 11–15 |
| Background or room change | 16–20 |
| Old-photo repair | 21–24 |
| Social-media version | 25–30 |
| Strong face preservation | Add the identity LOCK first |
| Strong product preservation | Lock label, logo, shape, colour and material |
| One small local change | Use the selection tool + written boundary |
| Several campaign versions | Return to the original for each major variation |
Frequently Asked Questions
Can ChatGPT Remove Someone From a Photo Without Changing Me?
Yes, but tell it explicitly that you must remain unchanged. For an important portrait, select the unwanted person where possible and name your face, hair, body, pose and clothing as locked details.
The selection tool can help narrow the edit, but OpenAI warns that the changed region can extend beyond the highlighted area. OpenAI image editor instructions
Can ChatGPT Make a Low-Quality Photo High Resolution?
It can make a photograph appear cleaner and more detailed, but generated detail should not automatically be treated as recovered information. Be especially careful with faces, text, products and old photographs where exactness matters.
Ask for conservative improvement.
Then compare.
Can ChatGPT Edit Product Photos for Shopify or Etsy?
Yes, it can be useful for background cleanup, lifestyle scenes, composition changes and campaign variations. Lock the product’s label, logo, colour, shape, material and proportions whenever the image is supposed to represent what the customer will receive.
If the package changes, regenerate from the source.
Do not sell Product A using AI’s accidental redesign of Product B.
Can ChatGPT Add Text to a Photo?
Yes. ChatGPT Images can generate text inside images, but for brand ads and social graphics I often prefer creating the visual space first and adding final copy afterward. OpenAI ChatGPT Images help
That gives you better control over spelling, font, line breaks, colour and placement.
The One Photo Editing Rule I’d Remember
You do not need to memorize all 30 prompts.
Remember this:
Tell ChatGPT what to change. Then tell it what it has no permission to touch.
If your face matters, lock it.
If the product matters, lock it.
If the room layout matters, lock it.
If the original photograph carries family history, treat reconstruction cautiously.
Then make one edit.
Compare it with the source.
Only after that would I move to the next change.
The best AI photo edit is not the one that looks the most dramatic.
It is the one where the change you wanted works—and everything you cared about still looks like itself.