Extend one side of a photo, or add space all around

If someone’s face is sitting too close to the edge of your canvas preview, you don’t have to accept it disappearing around the side. Sometimes the answer is a different crop. Sometimes you need a little more background. Here’s how to ask for that extra space, on one side or all four, and check that the people in your photograph still look like themselves.

The useful part is being specific. “Make this fit a canvas” leaves quite a lot to guess. “Add some background on the right, keep the people as they are, and leave the other three edges alone” is a much clearer job.

This guide is the practical companion to our guide to AI extension, canvas wrap and cropping. Keep that one handy for the difference between the visible front and the material going around the frame.

First, check what actually needs more room.

Upload your original to the print creator and look at the front and side guides. Is a head too close to the right edge? Are you losing shoes at the bottom? Or are you fitting a landscape photograph into a much squarer print?

Try moving the photograph and a different available shape before changing the file. If you like the original composition, a suitable plain or mirrored side treatment may also be worth considering where offered. Mirroring can look odd beside a face or a straight line, so check that too.

If the problem is only on the right, you may only need extra room on the right. Adding a large border to every edge makes the people smaller within the picture and gives you more generated scenery than you need.

Extra room for a new shape and extra room for the canvas sides are two different measurements. An A-series print has a different shape from a square photograph before you account for any wrap. Get the finished front size and the required edge allowance straight first. There isn’t one universal “add an inch” instruction that fits every size, depth and layout.

Start with the best file you have.

Use the original from the camera, phone or photographer. A screenshot of a WhatsApp photograph is a poor starting point if a better copy still exists. Our full-resolution sending guide explains how to get that file across.

Save a separate working copy. Don’t overwrite your original, and don’t crop away useful scenery before asking for more of it. Give the tool the background it needs to work from.

Extending one side in ChatGPT.

1. Open ChatGPT and attach the original photograph. Describe the change you want. Image editing is available across ChatGPT plans, with usage limits depending on your access.
2. Name the edge as you see it on screen. Say “the right-hand side when looking at the photograph”, rather than “her right”, which could mean the opposite side.
3. Describe what can be added. For example: more plain wall, grass, sky or the existing curtain. Ask it to continue the light, perspective and background texture.
4. Give an amount or target shape. “About a quarter of the original width on the right” is useful for a trial. If you know the required final dimensions, give those instead and check the downloaded result against them.
5. Name the things that must stay as they are. Faces, expression, hands, clothing, pet markings, colours and meaningful objects all matter. Ask for no extra sharpening or beautifying.
6. Download the result and compare it with the original. Don’t judge it only from the small image in the conversation. Return it to the print creator and check the front and sides again.

You can use the image selection tools where available, but a selection is not a guarantee that everything outside it will stay unchanged. OpenAI’s image-editing help explains that edits can extend beyond the selected area. A careful prompt helps; looking at the result is still part of the job.

Add room on one side

Attach the original and replace the bracketed parts. Left and right mean the edges you see on screen.

Edit this prompt for your photoExtend only the [right / left / top / bottom] edge of this uploaded photograph as it appears on screen. Add approximately [amount, for example 25% of the original width on the right]. Keep the other three edges in their original positions. Continue only the existing [grass / sky / wall / curtain / other background], matching its light, colour and perspective. Keep the people or pets, faces, expression, eyes, hands, clothing, distinctive markings, existing objects and original colours as they are. Do not add sharpening, smoothing, beautifying or a new photographic style. Do not invent missing people or body parts. If the requested area would require that, explain it before editing. Do not crop the original to make the new shape. My intended visible print is [width × height and units]; the confirmed allowance outside it is [allowance or not yet known]. Do not guess a wrap allowance if it is not provided. Return a separate edited file. State any known output pixel dimensions, without inventing metadata, and tell me if this workflow cannot guarantee preservation of the original image area.

Copy promptResetOpen ChatGPT ↗

Adding space all around the photograph.

This is useful when several edges are tight, or when you want more background around the whole subject. Ask for each edge separately: left, right, top and bottom. “20% on every edge” should mean 20% of the original width on the left and right, and 20% of the original height above and below. That makes the new dimensions 140% of the original in each direction.

That calculation describes the request, not a promise about the file a generator returns. Check the downloaded dimensions and the amount of room it has actually added. If the tool redraws or resizes the whole image, the numbers can be different.

More scenery also changes the composition. Your family will look smaller within a fixed-size print. That might be exactly what you need to keep everyone on the front, but it’s worth seeing the complete preview before deciding.

Add room on all four sides

Use the original again. These percentages describe the space requested, so still check the file you download.

Edit this prompt for your photoAdd background around all four edges of this original photograph. Add [20%] of the ORIGINAL width at the left and the same at the right; add [20%] of the ORIGINAL height at the top and the same at the bottom. Calculate the total dimensions from my chosen per-edge amounts: 100% plus twice the side percentage for width, and 100% plus twice the top/bottom percentage for height. Keep the original photograph centred, without cropping it or deliberately resizing its subject. Continue the existing background naturally. Preserve the people or pets, expression, facial features, hands, body proportions, clothing, markings, original objects, light and colours. Do not beautify, sharpen, restyle or redraw the subject. Do not add new people or meaningful objects. If an original edge cuts through a person, animal or meaningful object and extending it would require inventing the missing part, explain that before editing. I need extra background for [composition / confirmed canvas allowance], not an invented replacement for the photographed moment. Make a separate result. Tell me what output dimensions are actually available and whether the original image area can be guaranteed unchanged in this workflow. I will compare both files and check the print layout afterwards.

Copy promptResetOpen ChatGPT ↗

What happened with our examples?

We used the puppy photograph at the top of this page as the reference. We asked for extra grass on the right in one request, and extra space on all four edges in a separate request. Both started from the original, rather than editing the first result again.

These are AI editing examples, not photographs of a finished print or proof that the original pixels have been kept. The puppy’s body and toy already meet some of the original edges. That matters: extending those edges can require the tool to invent parts of them, not simply add empty grass.

The wider result gives us more lawn on the right, as requested. It also changes fine fur, the eye reflection and toy fibres. The original is 1455 × 1455 pixels; this returned file is 1402 × 1122. More background has not meant more pixels or an unchanged photograph.

The all-round result places the puppy further inside the picture. It has also invented the previously cropped body edge, left paw and claws, and a continuation of the toy. The returned file is 1254 × 1254 pixels. This is a useful example of why we would inspect the subject before accepting an extension for a personal print.

If a person’s arm, foot or head is already cut off in the original, AI cannot know exactly what was outside the photograph. It can make something plausible. For a meaningful family picture, asking the photographer for another frame may be a much better answer.

How much extra background do you need?

Use the actual print layout, not a guess from a social-media image ratio. As a simple example, a 12 × 8 inch front with an agreed allowance of 1 inch on every edge needs a 14 × 10 inch artwork area. At 200 pixels per inch that is 2800 × 2000 pixels. The 12 × 8 front sits inside it at 2400 × 1600 pixels.

That inch is an example, not our standard wrap specification. Frame depth, the turn around the back and the chosen side treatment can change what is required. Check the creator’s guides or ask us before making a file to a fixed measurement. Adding blank pixels or changing a DPI label doesn’t create photographic detail either.

For a calculation from your own file, use the calculator in our phone-photo print-size guide. Any generated margins also need checking for resolution and visible quality.

Check the people first. Then check the new background.

Put the original and the result side by side. Look at faces, eye shape, smiles, hands, pet markings and clothing. Check that the colours haven’t shifted. A lovely new background isn’t much help if the person in front of it has changed.

Then look along the join: repeated grass, a bent door frame, a changing horizon or a shadow going in the wrong direction can become more noticeable in a print. Check the full picture as well as close up. You want a photograph that works on your wall, not just a convincing thumbnail.

If the subject has changed, start again from the original. Don’t keep asking for corrections to an increasingly different face. This prompt helps you compare two files before deciding.

Check the extension before printing

Attach both files and identify which is the original. This asks for a comparison, not another edit.

Edit this prompt for your photoImage 1 is my original photograph. Image 2 is the extended result. Compare them without editing either file. First check whether faces, expression, eyes, hands, clothing, pet markings, meaningful objects and colours have changed. Then check which edges gained room, whether any original area was cropped, and whether the new background has repeated objects, odd seams, bent straight lines or inconsistent shadows. Separate a visible difference from anything you cannot judge because the files have different dimensions or display scales. Do not claim pixel-for-pixel equality from visual similarity or matching dimensions. Do not invent file metadata. My intended print is [size and units] with [confirmed allowance]. Tell me what I should inspect in the print creator, and whether returning to the original or choosing a different layout would be the better next step.

Copy promptResetOpen ChatGPT ↗

When keeping the original exactly matters.

If you need the original image area preserved, use an editor with layers: enlarge the document, keep the original photograph on its own locked layer, and make the new background on separate layers outside it. Avoid resizing or adjusting that original layer. Keep the layered working file so you can go back.

Photoshop’s Generative Expand is one route for extending a document with the Crop tool. Photoshop is a paid option, and generative access can use credits. It isn’t a free-tool recommendation. The layer setup and final comparison still matter.

A flattened export or colour conversion can change file values, so don’t treat matching file dimensions as proof that two files are identical. For most customers the practical question is whether the important people, colours and details have stayed faithful. If you’re unsure, get in touch before ordering and tell us the size you have in mind.

For changing the scenery behind a subject, rather than adding room outside the picture, see our background removal and replacement guide. For a small or soft source file, start with what’s worth trying with a free upscaler. They solve different problems, and you may not need every tool.

A few questions before you try it.

Can I extend just the left or right side of a photo?

Yes, that is a useful request to try. Name the side as it appears on screen, give an amount and describe the background to continue. Check that the tool has left the other edges and the subject as you wanted.

Can ChatGPT add space without changing faces?

You can ask for that, but a prompt does not guarantee unchanged faces or pixels. Compare the downloaded result with the original. If exact preservation is essential, use a layered workflow that keeps the original area separate.

Do I need to extend all four sides for canvas printing?

Only if the layout needs it. Check the visible front and side guides first. A one-sided extension, a different shape or another side treatment may be enough.

How much background should I add for a canvas wrap?

Use the allowance for the chosen print size, depth and layout. Don’t assume the frame depth alone is the complete allowance, or that a fixed percentage works for every canvas.

Does adding background improve a blurry face?

No. It adds room for the composition. Focus and source detail need a separate assessment, and any newly generated sharpness may include invented features.

Where we checked the editing steps.

We checked the linked OpenAI image-editing guidance and Adobe Generative Expand instructions on 24 October 2026. Menus and account allowances can change. The examples above are the returned edits from this guide; they aren’t a comparison of every editing tool.