These names are custom shortcuts for instructions, not built-in ChatGPT commands. The slash is a label. Paste a full definition below before using its nickname in the same conversation. A new chat needs the definition again; do not rely on ChatGPT remembering it across chats.
ChatGPT currently supports creating images and editing uploaded images on all tiers. You can describe the edit in conversation or select an image and highlight an area. A selection is not a perfect boundary: changes can extend outside it. Availability and usage limits can change, so use the limits shown in your account rather than assuming unlimited edits. OpenAI’s image guide explains the current controls.
Use a photo with a composition you already like. This is for light, colour and distracting noise, not a new face. A successful result looks like the same moment photographed more carefully.
For this conversation, /proshot means the following photo-editing instructions. Edit the attached photo into a natural, polished camera photograph. Preserve the person's identity, face shape, skin tone, age, expression, body proportions, hair, clothing and background composition. Improve exposure, white balance and local contrast gently; retain real skin texture and avoid smoothing, reshaping or invented makeup. Keep the existing light direction and use restrained sharpening rather than artificial crisp edges. Do not add objects, logos or text. Create one edited image with the original aspect ratio now, and use this definition when I ask for /proshot again in this chat.Try it on a softly lit phone portrait. Check that freckles, teeth and the shape of the eyes still match. If the face changes, start again from the original and request exposure and white balance changes only.
Use your best available scan, not a screenshot of a screenshot. Restoration can make an image easier to view, but generated detail is an interpretation. It cannot establish what a blurred face or damaged sign originally looked like.
For this conversation, /hdreal means a conservative restoration of the attached image. Reduce visible noise, mild blur and scanning damage while preserving the original composition, facial structure, expression, clothing and period details. Do not modernize the scene, colourize a monochrome original, invent readable lettering or add decorative detail. Where the source cannot support a detail, keep it soft rather than creating a confident new feature. Produce one restored image at the source aspect ratio. Afterward, briefly identify any areas that required interpretation so I can compare them with the original; do not claim to have recovered factual missing detail. Apply this definition now and whenever I use /hdreal in this conversation.Keep the original and restored version together. If a family member looks different, reject that version. For records where faithful pixels matter, use conventional restoration tools or a specialist instead of accepting a generated reconstruction.
Name a specific setting and its lighting. “Warm grey studio wall with soft light from camera left” gives you something much easier to judge than “better background.”
For this conversation, /newbg means replace only the background of my attached photo with the setting I describe. For this first edit, use a warm grey studio wall and floor with soft daylight from camera left. Preserve the subject's identity, pose, expression, body shape, clothing and accessories. Keep fine hair edges and semi-transparent details natural. Match the new background perspective and lighting to the subject, adding only the contact shadow needed to ground them. Do not replace the face, change the outfit or add furniture or text. Generate one image at the source aspect ratio. If my later /newbg request names another setting, use that setting with the same preservation rules.Look closely around hair, glasses and shoulders. A dark halo or inconsistent floor shadow is a reason to request a specific correction, not reroll the whole aesthetic.
Attach two images and identify them as Photo A and Photo B. Similar camera height and light direction make this easier. Treat the output as a created composite, not documentation of an event that happened.
For this conversation, /together means create a realistic composite using the two reference photos I attach. Photo A supplies the person who should stand on the left; Photo B supplies the person on the right. Preserve each person's distinct identity, age, face, hair, clothing and proportions without blending their features. Place them side by side with a comfortable gap against a simple neutral wall. Match camera height, scale, light direction and shadows; keep hands in natural positions and avoid invented physical contact. If either person is too obscured to use reliably, ask for a clearer reference before generating. Create one landscape image now. Use these same rules for later /together requests in this chat.Check each face separately against its reference. If one person drifts, try less ambitious posing and a simpler background before adding more instructions.
A clear front-facing image works better than a face hidden by hair. This is a visual comparison, not a guarantee that a cut will behave the same way with your hair texture or daily styling.
For this conversation, /hairstyles means create a four-panel haircut comparison using my attached portrait. Keep the same identity, facial features, expression, skin tone, camera angle, clothing and lighting in every panel. Show these four labelled options: 1 shoulder-length blunt cut, 2 long layers, 3 chin-length bob, 4 curtain fringe with the current length. Preserve my current hair colour and approximate natural texture; change only the haircut and styling needed for each option. Use equal-size panels and small readable number labels, with no beauty retouching or claims about which cut objectively suits me. Generate the comparison now. For later /hairstyles requests, replace the four options with the ones I specify.If the grid keeps changing your face, request one option at a time with the original reference. Show the options to your stylist to discuss what is practical for your actual hair.
Use this to narrow a style direction. It does not measure bridge width, temple length, prescription compatibility or comfort.
For this conversation, /sunglassestryon means create a four-panel sunglasses style comparison from my attached front-facing portrait. Preserve the same face, skin tone, expression, hair, clothing and camera angle across all panels. Show four unbranded frame shapes: rectangular black acetate, thin round metal, subtle cat-eye acetate and classic aviator. Keep each frame plausibly positioned at the bridge and temples with realistic shadows and restrained lens reflections. Use the same neutral background, equal-size panels and small labels 1 to 4. Do not alter my facial proportions to accommodate a frame or present this as a physical sizing test. Create the grid now and use this definition for future requests in this conversation.Check the bridge, the arms beside your head and whether your face changed. Once you like a shape, check the real product measurements or try it on before buying.
A useful shortcut names an action, states what must stay unchanged and defines a result you can check. Use the builder below with a repeated task, such as preparing a profile photo with a consistent crop and background.
Help me create a reusable photo-editing shortcut for this task: prepare a square profile photo with a warm grey background and room around my head. Give it a short memorable name beginning with /, but explain that it is a custom instruction label for this conversation, not a built-in command. Write a complete reusable definition that specifies the reference input, exact changes, identity and clothing preservation rules, output composition and three visual acceptance checks. Ask up to three questions only if essential details are missing; otherwise state sensible defaults. Return the finished definition in one copyable block and one example invocation. Do not edit an image until I attach a reference and ask you to apply the definition.Save the definition in your notes, not only the nickname. For your next task, replace the first sentence with the edit you repeat most often.
If ChatGPT replies with advice instead of an image, ask it to create the edited image using the attached reference, or open Images and paste the full prompt there. If a limit appears, follow the account message rather than repeatedly submitting the same request.
Change one thing per revision: “Restore the original eye shape; keep this version's background” is more useful than “make it realistic.” For sensitive identity details, return to the original reference instead of building a chain of increasingly altered faces. Use pictures you have permission to edit, and keep a result only when it still represents what you intended.
Choose one clear photo, paste /proshot's full definition and compare the result side by side. Once that works, save the definition and reuse it with a new reference. The time-saving part is a tested instruction you can repeat.