Photo editing guide
ChatGPT photo editor: how to edit photos with prompts
To edit an existing photo with ChatGPT, upload the image, describe the change, review the result, and refine it with a follow-up prompt. You can also open an image in ChatGPT's editor and select an area for a more targeted change.
The short answer
Can ChatGPT edit an existing photo?
Yes. ChatGPT Images accepts an uploaded photo or an image made in ChatGPT. You can request an edit in plain language, or select part of an image and describe what to change there. OpenAI says the image editor can add, remove, and update parts of an image. Its selection is a guide rather than a pixel-perfect mask: a change can extend beyond the highlighted area.
An edit is a generated result, so inspect faces, hands, product labels, text, and other details that must remain accurate. Keep the original file until you have checked and saved the version you want.
Choose the right starting point
ChatGPT image generation vs. photo editing
When you ask ChatGPT to create an image from text, there is no source photo to preserve. When you upload a photo or open an image already generated in ChatGPT, you can ask for a change to that image. If a real person's likeness, a product's shape, or a scene's composition matters, attach the source image and say which details should stay. Check the output rather than assuming it retained them exactly.
Step by step
How to edit a photo with ChatGPT
Upload the photo
In a ChatGPT conversation, use the attachment control to add a photo. You can also drag an image into the prompt area or paste a copied image. If you already generated an image in ChatGPT, open that image instead.
Describe one specific edit
Name the subject or area, the change you want, and what must stay the same. “Replace the gray wall behind the person with a plain warm-white wall; keep the person, clothing, pose, and crop” is more actionable than “make it better.”
Select an area when the edit is local
Open the image editor and use Select to highlight the object or region. Then describe the change in the text box. You can also skip selection and describe the location in words, such as “the lamp on the left.”
Review and refine
Compare the result with the original at full size. If something changed unexpectedly, send a short correction that names the problem. For a new direction, go back to the original and try a separate prompt.
Save the result
Open the image controls and choose Save to download the finished image. Keep the original separately so you can compare or restart later. OpenAI also places images created in ChatGPT under Images for browsing and reuse.
Better instructions
How to write a ChatGPT photo-editing prompt
Use a short instruction with four parts: identify the area, state the change, name the details to preserve, and specify the output if it matters. A preservation instruction is useful guidance, though it cannot guarantee that every detail will remain identical.
Edit [specific area] by [specific change]. Keep [subject, identity, pose, framing, colors, or text that must stay]. Make the result [realistic style or output requirement].
For example: “Brighten the face slightly and reduce the cool color cast. Keep the facial features, skin texture, hair, clothes, background, and crop. Make it look like a natural photograph.” Ask for one major change at a time; after reviewing the result, add the next edit.
Examples by task
ChatGPT photo-editing prompts you can adapt
Replace the bracketed details with what is actually in your photo. Each prompt identifies the part to change and the elements to check afterward.
Remove a background
Remove the background behind [subject] and make it transparent. Keep the entire subject visible, including hair and fine edges. Do not change the subject's appearance.
Check the outline and confirm that the saved file has transparency.
Change a photo background
Replace only the background with [setting]. Match its lighting and perspective to the subject. Keep the person's face, clothing, pose, and framing as close to the original as possible.
Check that the new setting has believable shadows and scale.
Remove an object
Remove the [object] in the [location]. Fill the space with the surrounding [wall/ground/sky] texture. Keep the main subject and the rest of the composition unchanged.
Inspect the repaired area for repeated patterns or distorted edges.
Retouch a portrait
Reduce temporary blemishes and soften harsh shine. Preserve the person's facial structure, expression, skin texture, hairstyle, and natural features. Avoid a smoothed or plastic look.
Compare eyes, teeth, hands, and facial proportions with the source.
Fix lighting and color
Lift the shadows slightly and neutralize the blue cast. Keep the scene, subject, and framing the same. Preserve natural skin tones and avoid clipped highlights.
Check bright areas and color consistency across the image.
Improve a product photo
Place this [product] against a clean [background]. Keep its shape, color, logo, label text, and packaging details accurate. Add a soft, realistic contact shadow.
Verify every printed word and brand detail before using the result.
Restore an old photograph
Reduce scratches, dust, and fading in this photo. Preserve the original faces, clothing, composition, and period details. Do not invent missing features where the source is unclear.
Compare the restoration with the scan, especially around faces.
Resize or reframe
Reframe this photo for a [square/portrait/wide] composition. Keep [subject] fully visible and near [position]. Extend only the surrounding background where needed.
Inspect newly generated edges; they may contain invented content.
Add or change text
Add the exact text “[words]” to the [location] in a clear [style]. Keep the rest of the image unchanged and make the text readable at its final size.
Read every character in the output before publishing it.
Apply a visual style
Give this photo a [style] color treatment with [lighting description]. Keep the people, objects, pose, and composition recognizable. Avoid adding new subjects or props.
Use the original image for another style direction if the result drifts.
Use a reference image
Use the first image as the photo to edit and the second only as a reference for [lighting/color/style]. Keep the first image's subject, pose, and composition. Do not copy people or objects from the reference.
Label each attachment's role clearly and verify the edited source.
Quality control
How to keep faces and important details consistent
A request to “keep the face exactly the same” tells the editor your priority, but it is not a guarantee. For identity-sensitive portraits, legal documents, product labels, or anything that must be pixel-accurate, compare the output carefully with the source and use a conventional image editor when exact preservation is required.
- Limit the scope. Change one area or one property per prompt.
- Name what must stay. Call out the face, expression, pose, garment, logo, text, or background explicitly.
- Select a local area. Selection can help, but OpenAI notes that edits may reach beyond it.
- Inspect at full size. Zoom in on fingers, teeth, edges, typography, and any repeated pattern.
- Keep the original. Start again from it when several follow-up edits have changed too much.
Troubleshooting
What to do when ChatGPT edits the wrong thing
The result looks like a new person or product
Return to the original. Request a smaller change and list the identity, shape, label, and framing details to preserve. Check the next result against the source.
A selected edit changes nearby areas
Use a smaller selection and name the exact object or region in the prompt. Inspect outside the selection too; OpenAI says the highlight is not always precise.
Text or a logo is incorrect
Provide the exact wording, request legible text, and verify the output character by character. Use a dedicated design editor for copy that must be exact.
The response describes the photo instead of editing it
State the desired output directly: “Create an edited version of this uploaded photo” followed by one concrete change. Confirm the image attachment is present before sending.
Where it works
ChatGPT photo editing on web, iPhone, and Android
OpenAI lists ChatGPT Images on the web, iOS, and Android, with image creation and editing available across its tiers. On mobile, open an image, choose Edit to describe a change, or use Select to highlight an area before sending the instruction. Exact controls and usage limits can change, so the controls on your screen may differ.
About GPTPhotoEditor
A photo-focused workspace with ChatGPT and Gemini image models
GPTPhotoEditor offers a ChatGPT image model alongside faster Gemini options in a workspace built around photo editing. Attach a photo, describe a change, and keep one current image in view. Revisit earlier versions in the thumbnail timeline, branch from any version to try a different direction, and download the result you prefer. Additional images and documents can travel with a prompt as references.
Create & chat guide
How to create an image with a prompt and refine the idea in chat
Choose Create & chat when you are starting with an idea or want to explore a photo in conversation. If you type a prompt in Edit photo without a source image, Send moves the request to Create & chat automatically. Switch between chat and Edit photo within the same session: image results in chat become versions in the editor, and the selected editor image appears in chat.
Write a visual brief
Name the subject, setting, composition, lighting, style, and any text that must appear. Include the intended shape, such as square or portrait, when it matters. For example: “Create a square product image of a matte black bottle on pale stone, with soft window light from the left, a natural contact shadow, and no text.”
Attach references with distinct roles
Use Attach or drop files onto the page to add multiple images, PDFs, or documents. Tell the conversation what each file is for: “Use the first image for the bottle shape and the second for the lighting. Use the brief for the color palette.” This helps separate the subject from the style reference.
Refine one decision at a time
Keep follow-up prompts in the same conversation when they build on the same idea. Say what should change and what should stay: “Keep the bottle and camera angle; make the background warmer and the shadow softer.” Choose New chat for a different concept instead of stacking conflicting directions.
Continue in Edit photo
Switch to Edit photo to work on the latest image result in the same session. Use View in editor on an earlier chat result to select that version. The timeline lets you compare versions and branch from a previous result without removing later versions from the original session.
Questions and answers
ChatGPT photo editor FAQ
Can ChatGPT edit a photo I upload?
Yes. Upload an existing image and describe the change you want. You can also edit an image generated in ChatGPT.
Can I edit only one part of the image?
Yes. Open the image editor, select the area, and describe the edit. The selected boundary is approximate, so inspect the rest of the image.
Can ChatGPT remove or replace a background?
Yes. Request a transparent background or describe the replacement setting. Review the subject's outline and the saved result.
Will ChatGPT keep a face exactly unchanged?
No prompt can guarantee exact preservation. Name the features to keep and compare the output with the original; use pixel-based tools if exactness is essential.
Is ChatGPT photo editing free?
OpenAI says ChatGPT Images is available across all tiers. Access, limits, and available controls may differ by plan or change over time; check OpenAI's current plan information.
How do I download the edited photo?
Open the finished image and use its Save control to download it to your device.
Is image generation the same as editing?
No. Generation starts from a text description; editing starts from an existing uploaded or generated image and asks for a change to it.
When should I use Create & chat instead of Edit photo?
Use Create & chat to explore image ideas and references as a conversation. Switch to Edit photo for focused changes to the current image; both views share one session and its image history.
Can I attach several references to one prompt?
Yes. Add multiple images or supported documents, then identify the role of each in your prompt so the subject, style, and written brief are clear.
Does GPTPhotoEditor offer a ChatGPT image model?
Yes. Choose Pro for the OpenAI image model, or use the Lite and Standard Gemini models. Each runs in the same editing workspace with a canvas, version history, branching, and reference attachments.