The ChatGPT Image Codes Cheat Sheet
Those slash codes aren’t secret OpenAI settings. They’re shorthand for image instructions. Here’s how to pick a look without letting ChatGPT make every other decision.
TL;DR
ChatGPT image codes are community shorthand, not official hidden commands. Treat these 242 entries as a menu of intended looks, not guaranteed effects. Upload your photo, pick a code, and explain where you’ll use the picture, what must stay unchanged, and the crop. Compare the result with your original, not just your idea of a polished photo. If it stops looking like you, restart from the original instead of editing a face that’s already wrong.
Key takeaways
- The slash isn’t a switch. ChatGPT interprets the words as instructions, so unfamiliar shorthand may need a plain-English explanation.
- You might mean better lighting by “professional.” ChatGPT might also change your clothes or face unless you tell it what needs to stay.
- Pick a code, then tell ChatGPT where you’ll use the picture, what can change, what needs to stay, and how to frame it.
- Combine requests that work together, such as /proshot and /bgstudio. Two or three codes is a practical stopping point, not a technical limit.
- Sharper-looking detail isn’t necessarily recovered detail. Check your likeness, hands, text, crop, and actual file dimensions before publishing.
- The cheat sheet contains 242 entries in 18 groups. The source doesn’t explain /dat, so write out your intended scene instead of guessing.
You need a better headshot, but the polished version doesn’t quite look like you. The lighting improved. Apparently, your jaw needed a promotion too.
Then someone sends you a list of ChatGPT image codes: /proshot for a professional photo, /lego for a minifigure. They look like shortcuts to predictable results.
These aren’t official hidden ChatGPT commands. They’re community shorthand for longer image instructions. ChatGPT interprets the words; the slash doesn’t trigger a fixed editing operation.
Think of the list as a menu, not a control panel. You’ll find all 242 entries in 18 groups below. First, let’s make sure you get the picture you wanted, not just a nicer picture of someone suspiciously similar.
The code names a look. You still need to explain the job.
A prompt is the instruction you give an AI tool such as ChatGPT. An image code shortens that instruction, but it leaves plenty unsaid.
Take /proshot. You could mean a LinkedIn headshot, a fashion portrait, or a full-body business photo. The code doesn’t choose between them, and it doesn’t say which details need to stay unchanged.
That’s the hidden assumption: permission to improve the photo isn’t permission to improve the person. You may want better lighting. ChatGPT may also change your skin, hair, clothes, or apparent age.
Before you pick a code, think about what would make the picture useful. Your colleagues should recognize your headshot without help from the caption. That gives you a simple way to explain what you want:
- Keep this: Your facial features, apparent age, build, hairstyle, and clothing.
- Change this: The uneven lighting or distracting background.
- You can choose: A subtle background shade.
Now you have something better than “I’ll know it when I see it.” If the lighting looks better but the face looks wrong, you know what to keep and what to fix.
Start with one photo and one focused request
You’ll get a more useful starting point from a clear photo you have permission to upload and edit. It should show the features you want to keep. If your face is blurry or partly hidden, you’re asking ChatGPT to fill in details it can’t see.
Pick a code, then tell ChatGPT where you’ll use the picture, what can change, and what needs to stay. Include the framing, too. For example: “Keep my face recognizable; make this vertical for Instagram, with room above my head.”
For a LinkedIn headshot, you can copy the fuller request below. Change the destination and framing to fit your needs.
Use /proshot as shorthand for a professional-looking photo, not permission to redesign my appearance.Create a realistic headshot from the uploaded photo for my LinkedIn profile.
Keep my facial features, apparent age, build, hairstyle, expression, and clothing unchanged. Preserve natural skin texture.
Improve the lighting toward neutral daylight and use a simple studio background. You may choose a subtle background shade.
Frame me vertically from the chest up, leaving enough space for a square profile crop without cutting off my head.
If an important detail is unclear in the source, ask rather than inventing it. Ask one question at a time and wait for my answer.
You’re giving ChatGPT less room to guess, not a guarantee that every detail will stay put. Keep the original handy so you can compare them.
Combine requests when they help each other
These two work together: /proshot asks for a professional-looking photo, and /bgstudio asks for a studio background. If you’re curious about what each changes, try them separately first.
Two or three codes is a practical stopping point, not a technical limit. The trouble starts when your requests disagree, not when you reach a certain number of slashes.
For example, /keepitreal and /doll pull in different directions. More instructions won’t settle an argument between the first two.
If the result misses, another code usually tells ChatGPT less than a sentence about what went wrong:
The lighting is too warm. Make it neutral daylight, keeping my face, clothes, and framing unchanged.
With that in mind, browse the list for the look or change you need. You don’t need to memorize it. You just need a starting point you can explain.
The complete ChatGPT image codes cheat sheet
These 242 entries are grouped by purpose. The descriptions explain intended requests, not guaranteed behavior. If ChatGPT misreads a code, use its description and say what needs to stay unchanged.
Some codes name a look; others suggest a setting, pose, or format. Few cover everything you need. “Ready” and “HD,” meaning high definition, aren’t quality certificates.
Portraits and face fixes
Start here for headshots and everyday cleanup. “Correct the uneven lighting” gives ChatGPT a smaller job than “make me look better.” That matters when you like your face as issued.
/portrait— A polished, realistic portrait from your photo./hdportrait— A highly detailed, high-resolution portrait; check the actual file dimensions./storyportrait— A portrait with a stronger story or emotional context./natureportrait— A portrait in a natural outdoor setting./proshot— A professional-quality photographic look./perfectshot— Improvements toward an ideal camera capture./cameraready— A polished, camera-ready appearance./photofix— Corrections to lighting, color, sharpness, and other photo issues./bestlook— A flattering but recognizable version of you./glowup— A tasteful visual upgrade to your appearance./makeitbetter— Overall improvement while preserving the image’s core look./photorescue— Restoration or substantial improvement of a poor-quality photo./keepitreal— Improvements that keep the image realistic./instafix— An Instagram-ready appearance./dpready— A polished profile picture./selfiefix— A more polished selfie that still looks natural./facefix— Corrections to facial inconsistencies or imperfections; specify what you mean./faceglow— Healthy-looking light and vibrancy in the face./skinclean— Skin cleanup without dramatically changing your appearance./skinreal— Realistic skin texture rather than an airbrushed finish./eyepop— Clearer, more visually engaging eyes./smilefix— A natural-looking adjustment to an awkward smile./makeover— A more significant style makeover./hairfix— Hairstyle corrections or stray-hair cleanup./hairchange— A different hairstyle while preserving the person’s likeness./haircolor— A realistic hair-color change./beard— Added or modified facial hair./glasses— Realistic glasses added to the subject.
Other versions of you
Here you’re moving beyond cleanup and asking ChatGPT to invent something. Enjoy the result, but don’t treat it as a prediction or recovered memory. A believable childhood portrait doesn’t tell you what you actually looked like.
/minime— A tiny miniature version of you./giantme— An exaggerated giant version of you./cloneme— Multiple copies of you in the same scene./digitaltwin— A realistic computer-generated visual duplicate of you./childme— You reimagined as a child./childhoodme— A believable younger childhood version of you./futureme— An imagined future appearance./thennow— Younger and current versions of you together./old2new— An older photograph modernized or refreshed./myera— You in a specified historical or cultural era./altlife— You living an alternate version of your life./parallelme— Alternate-universe versions of you./timesplit— Versions of you from different time periods together./portalme— You entering or emerging from another world or era./dreamlife— You living an aspirational dream scenario./maincharacter— You as the central character in a movie-like scene./movieversion— You reimagined as a feature-film character./paparazzi— A candid celebrity-style paparazzi photograph./redcarpet— You at a glamorous red-carpet event./billboardme— Your image placed realistically on a large billboard.
Toys, models and stylized figures
/lego— A LEGO-style minifigure or scene./toy— A person or object turned into a collectible toy./actionfigure— A packaged action-figure version of the subject./miniature— A scene or person turned into a tiny model./clay— A clay-style figure./statue— The subject as a sculpted statue./doll— A stylized doll version of the subject./papercraft— A handcrafted paper-model version./ghibli— Whimsical, hand-painted animation qualities.
Two or more people in one photo
If you’re combining people, tell ChatGPT how they know each other and what they’re doing. “Together” doesn’t say whether they’re colleagues, siblings, or a couple.
Before you create or share an invented affectionate scene, get permission from the real people involved. A convincing picture doesn’t establish that the moment happened.
/together— People from separate photos combined into one believable scene./groupshot— A natural-looking group photograph./coupleportrait— A polished portrait of two people together./candidcouple— A natural, unposed photo of a couple./coupleselfie— A believable selfie of two people together./face2face— Two people positioned closely, facing one another./backhug— One person hugging another from behind./hug— A natural embrace between people./sidehug— People embracing from the side./closehug— A warmer, tighter embrace./sittingtogether— People seated naturally beside one another./walkingtogether— People walking naturally together./walkinghandinhand— Two people walking while holding hands./headonshoulder— One person’s head resting on another’s shoulder./headtohead— Two people’s heads positioned affectionately together./foreheadkiss— A gentle kiss on the forehead./cheekkiss— A natural kiss on the cheek./couplegaze— Two people looking naturally at each other./playfulcouple— A lighthearted, playful interaction./naturalcouple— An authentic-looking everyday photo of a couple./cinematiccouple— A dramatic, movie-like image of two people.
Outfits and style
/outfit— Changed or newly created clothing./outfitup— A more polished or stylish outfit./outfitfix— Corrections to awkward or inconsistent clothing details./outfitmatch— Coordinated clothing across multiple people./styleup— A more stylish overall appearance./newlook— A noticeably different fashion or personal style./oldmoney— A classic, understated luxury look./quietluxury— An elegant, refined, logo-free luxury look./casualstyle— A relaxed everyday outfit./formalstyle— Formal professional or evening clothing./datelook— A polished outfit suitable for a date./partylook— A stylish party outfit./weddinglook— Wedding-appropriate clothing and styling.
Camera angles and framing
A new angle may show parts of you that aren’t visible in the original photo. ChatGPT has to imagine those details, not recover them. Keep that in mind if you need the picture to look accurately like you.
/closeup— Tight framing, usually around the face./macro— An extremely close, highly detailed camera view./sideview— A view primarily from the side./backview— A view primarily from behind./topview— A camera angle looking down from above./lowangle— An upward-looking angle for a more powerful appearance./wideangle— A broader scene with a wide-angle perspective./droneview— A simulated high aerial drone photograph./birdeyeview— A directly overhead viewpoint./povshot— The scene viewed through a person’s eyes./overtheshoulder— Framing from behind someone’s shoulder.
Photographic effects
/bokeh— Soft, out-of-focus background lights and blur./motionblur— Movement blur suggesting speed or action./longexposure— The look of a slow camera exposure, such as light trails or smoothed moving water./filmgrain— Subtle speckled texture resembling photographic film./polaroid— The character of an instant-film photograph./doubleexposure— Two images blended into one overlapping composition./silhouette— The subject as a dramatic dark outline./reflection— Added or emphasized realistic reflections./lensflare— Streaks or glows from a bright light source.
Lighting
/goldenhour— Warm sunrise or sunset lighting./softlight— Gentle, flattering light with soft shadows./windowlight— Natural-looking illumination from a window./lightfix— Corrections to poor or uneven lighting./nightfix— Improvements to a dark nighttime photograph./darktobright— A brighter image with realistic-looking detail./rembrandt— Classic dramatic portrait lighting with facial shadow./godrays— Visible, dramatic beams of light through the scene./practicallights— Visible lamps or fixtures used as movie-like light sources./twotonelight— Two contrasting light sources for a movie-like look.
Detail, sharpness and color
Sharper doesn’t necessarily mean more accurate. A cleanup can invent convincing teeth, eyes, or lettering where the source is unclear. Compare those details with the original, even when the new version looks better.
/hdreal— More realistic-looking detail and photographic clarity./detailup— Additional fine detail throughout the image./sharpreal— A sharper image without an artificial-looking finish./blurfix— Reduced unwanted blur where possible./colorfix— Corrections to inaccurate or unattractive colors.
Backgrounds
/newbg— A replacement background./bgfix— Repairs or improvements to the existing background./bgremove— Removal of the background around the subject./bgpro— A polished professional background./bgnature— A natural outdoor background./bgcity— A believable city environment./bgstudio— Professional studio scenery./bgcinematic— A dramatic movie-style background./bgluxury— An upscale luxury environment./bgblur— A blurred background with a sharp subject./cleanbackground— A simpler background without distracting elements./messyclean— Clutter removed from a messy environment.
Removing people and objects
/personremove— Removal of a specified unwanted person./peopleclean— Removal of distracting background people./crowdremove— Removal or substantial reduction of a crowd./objectremove— Removal of a specified unwanted object./distractionremove— Removal of miscellaneous visual distractions.
Places and environments
/foggyforest— An atmospheric foggy forest./mistymountain— A mist-covered mountain environment./rainyforest— A wet, atmospheric forest scene./snowyforest— A snow-covered forest./sunsetbeach— A beach scene during sunset./moonlight— Dramatic moonlit illumination./underwater— The subject or scene reimagined underwater./space— An outer-space environment./jungle— A lush jungle setting./desert— A dramatic desert landscape.
Social-ready formats
You’ll still need to say where you’re posting and which crop you need. “Ready” in a code name doesn’t tell ChatGPT where the picture will go. It also doesn’t promise that the downloaded file will fit.
/instaready— An image prepared for posting on Instagram./storyready— Formatting and styling for a social Story./postready— A polished social-media image./reelcover— A visually strong vertical Reel cover./aesthetic— A more intentional, cohesive visual style./mirrorselfie— A believable mirror-selfie composition./candidlook— A natural, unposed photographic appearance./streetsnap— A spontaneous street-photography look.
Birthdays and celebrations
/birthdaywish— A personalized birthday greeting image./birthdayposter— A celebratory birthday poster./birthdaycard— A personalized birthday card design./birthdaystory— A birthday-themed social Story image./birthdayparty— A festive birthday-party setting./birthdayqueen— An exaggerated birthday-queen portrait./birthdayking— An exaggerated birthday-king portrait./surpriseparty— A lively surprise-party scene.
Couples, closeness and candid moments
These codes give you more specific poses and settings for two people. As with the earlier group, explain their relationship. Get their agreement before inventing an affectionate moment involving them.
/couplefunselfie— A playful selfie of two people together./mirrorcoupleselfie— A couple taking a mirror selfie together./romanticselfie— A warm romantic selfie./cozyclosecouple— A comfortable close-up image of a couple./intimatefaceclose— An emotionally close facial composition./faceclose— Two people’s faces positioned close together./romanticcloseup— A close-up romantic portrait./waisthug— One person hugging another around the waist./shoulderhug— An embrace around the shoulders./protectivehug— A warm, protective-looking embrace./gentlehug— A soft, understated embrace./shoulderlean— One person leaning naturally onto another’s shoulder./chinrestcouple— One person resting their chin affectionately near the other./behindshoulder— One person positioned just behind another’s shoulder./cheektohead— One person’s cheek resting gently against another’s head./dat— No description was supplied in the source list. Write out your intended scene instead./handholdcouple— A natural scene of a couple holding hands./nosekiss— A playful, affectionate nose-to-nose moment./almostkiss— The moment immediately before a kiss./foreheadtouch— Two people’s foreheads gently touching./eyelookcouple— Emphasized eye contact between two people./romanticlook— A subtle romantic exchange of expressions./lovinggaze— Two people looking at each other affectionately./cuteplayful— A cute, lighthearted interaction./softromance— A subtle, understated romantic atmosphere./cozyromance— A warm, comfortable romantic environment./datecouple— A couple spending time naturally on a date./cafedate— A couple together on a café date./cozycafedate— A warm, intimate café setting for two./coffeecouple— Two people enjoying coffee together./picniccouple— A couple on a picturesque picnic./sunsetcouple— A couple portrait during sunset./goldenhourcouple— A couple in warm sunrise or sunset light./raincouple— A movie-like scene of a couple in the rain./wintercouple— A couple in a cozy winter environment./beachcouple— A couple together at the beach./mountaincouple— A couple in a mountain setting./streetcouple— A natural street-style couple photograph./rooftopcouple— A couple on an urban rooftop./balconycouple— A couple together on a balcony./windowcouple— A couple portrait beside a window./cozyhomecouple— A relaxed couple scene at home./carcouple— A couple together in or around a car./roadtripcouple— A couple enjoying a road-trip scene./travelcouple— A couple traveling together./royalcouple— A couple with regal clothing and surroundings./luxurycouple— A couple in a sophisticated luxury environment./editorialcouple— A couple portrait styled like a fashion-magazine shoot./filmlookcouple— A couple portrait with the appearance of photographic film./dreamycouple— A soft, dreamlike couple portrait./closecoupleportrait— A tightly framed professional couple portrait.
Art, print and film looks
/gouache— A painting with dense, matte color rather than transparent watercolor washes./linocut— A bold print resembling carved shapes stamped in ink./papercut— A scene made from layered cut-paper artwork./conceptart— Polished artwork exploring a character or setting for entertainment, such as a film or game./cyanotype— A classic blue-and-white photographic-print look./risoprint— A textured print look using a limited set of layered ink colors./cinestill— The intended request here is an image resembling a feature-film frame./lightleak— Colored glows resembling light accidentally reaching photographic film./bleachbypass— A high-contrast film look with muted colors./splittone— Different color tones applied to bright and shadowed areas./tealandorange— A movie-style color treatment pairing teal tones with orange warmth.
Design eras and poster styles
/bauhaus— A composition using simple geometric forms and restrained design./y2k— A late-1990s or early-2000s visual style./sovietposter— A vintage propaganda-poster composition./swissstyle— Clean, grid-based graphic design with orderly text and spacing.
Advanced composition
/negativespace— Intentional empty space around the subject./wormseye— An extreme upward-looking camera angle from ground level./framedwithin— Objects within the scene used to frame the subject./tightcrop— Close cropping around the primary subject./anamorphic— A widescreen, movie-lens appearance./tiltshift— A narrow band of focus, often used to make scenes resemble miniatures./shallowdepth— A sharp subject against a strongly blurred background.
Look for what changed without permission
Once you’ve tried a code, put the result beside your original. “Does this look good?” is tempting, but it’s not the most useful question. Ask, “Did it change only what I wanted changed?”
Suppose your headshot has better lighting, different teeth, and a clean background. The lighting and background worked, but the teeth changed—so asking it to “make it better” again doesn’t explain what went wrong.
When the mismatch is small, you can ask for that one correction while keeping the parts you like. If the background is the problem, for example:
Keep the face, clothing, and framing unchanged. Replace only the background with plain gray.
If the person no longer looks like you, restart from the original photo. Otherwise, repeated edits can carry the wrong features forward. Don’t let a mistake become the new reference.
Sometimes you don’t need ChatGPT to redraw anything
These codes are useful when you want to try a different look, make something playful, or explore how a portrait could look. They’re less useful when every original detail needs to stay exactly where it is.
For evidence, identity documents, or faithful historical restoration, don’t rely on a version created by AI. Looking convincing isn’t the same as being a faithful record.
If you just need a tighter crop or a less yellow photo, a regular photo editor can help. It can make those changes without taking another guess at your face.
Check the file, not just the preview
If the result looks good beside your original, download it before you publish. The preview won’t settle every question, so give these four areas a closer look:
- Does it still look like you? Compare eyes, teeth, face shape, skin texture, and apparent age with the source. A plausible face isn’t necessarily your face.
- Did the small details survive? Check hands, glasses, clothing edges, reflections, and text. A beautiful birthday card addressed to the wrong person has one fairly significant design problem.
- Does the file fit the job? Confirm the actual dimensions and crop. If you asked for background removal, check that the file has the transparency you need.
- What will people think happened? Confirm you have permission. Consider labeling invented scenes if viewers could mistake them for real events or relationships.
Your next move: give one code a clear job
Pick one photo and one change you actually need. If it’s a headshot, upload the original and copy the prompt above. Adjust the destination and framing, then try it once.
When it comes back, ask yourself two questions: does it still look like you, and does it fit where you’ll use it?
If the code isn’t helping, delete it and keep the plain-English instruction. The useful part was always your direction, not the slash.
Copy this prompt
Click copy, then paste it into ChatGPT (or any AI chat) and fill in the brackets.
Use /proshot as shorthand for a professional-looking photo, not permission to redesign my appearance.Create a realistic headshot from the uploaded photo for my LinkedIn profile.
Keep my facial features, apparent age, build, hairstyle, expression, and clothing unchanged. Preserve natural skin texture.
Improve the lighting toward neutral daylight and use a simple studio background. You may choose a subtle background shade.
Frame me vertically from the chest up, leaving enough space for a square profile crop without cutting off my head.
If an important detail is unclear in the source, ask rather than inventing it. Ask one question at a time and wait for my answer.
Step-by-step
1. Give the image a job
Think about where you’ll use it and what would make it unusable. A LinkedIn headshot won’t help if it stops looking like you or cuts off your head.
2. Decide what can change
Tell ChatGPT what needs to stay, what you want changed, and what it can choose. For a headshot, you might keep your features, ask for better lighting, and let it choose a subtle background shade.
3. Upload a clear photo
Choose a photo you have permission to upload and edit. It needs to show your face and any other details you want kept.
4. Ask for one focused change
Pick a code or write the request in plain language. Explain where you’ll use the picture, what can change, what needs to stay, and the crop.
5. Compare it with the original
Look for changes you didn’t request. Check the lighting, face, clothing, background, and framing rather than accepting the whole image because it looks polished.
6. Explain what needs fixing
Ask for a specific correction and name what should stay unchanged. If the face no longer looks like you, restart from the original photo.
7. Check the downloaded file
Before publishing, check facial features, hands, text, edges, reflections, crop, and actual dimensions. Check transparency if requested, and confirm permission. Consider whether an invented scene needs a label.
Frequently asked questions
Are ChatGPT image slash codes official commands?
No. They’re community shorthand for longer image instructions, not official hidden ChatGPT commands. ChatGPT interprets the words; the slash doesn’t run a fixed editing operation.
Why do ChatGPT image codes sometimes work well?
Recognizable words give ChatGPT a direction. For example, /bgstudio suggests a studio background. But that label doesn’t tell it where you’ll use the picture, what makes you look like you, or which crop you need. You still need to add those details.
Can I combine codes?
Yes, when the requests fit together. /proshot asks for a professional-looking photo, and /bgstudio asks for a studio background. Start with one code and add another when it supplies something missing. Two or three is a useful stopping point, not a technical limit.
What should I do when a code doesn’t work?
Replace the shorthand with a specific instruction. For example: “The lighting is too warm. Make it neutral daylight, keeping my face, clothes, and framing unchanged.” Tell it what went wrong instead of adding more codes.
Why does the result sometimes stop looking like me?
Broad appearance requests, conflicting instructions, and repeated edits can pull the result away from your original features. Compare both images side by side. If the face no longer looks like you, restart from the original with one narrower request.
Do detail and resolution codes recover missing information?
Don’t assume they do. A sharper-looking face may contain invented details. Asking for high resolution doesn’t guarantee particular file dimensions, either. Check the downloaded image. When you need the original details left alone, use a regular photo editor.
The Action Guide
Ready to put this into practice?
The article built the understanding. The Action Guide is where you actually do it — try it on your own work, and build the skill.
Open the Action Guide