One photo has to fit a story, a feed post and a banner: how to change its shape without losing the subject

A photo has one shape, and every platform wants a different one. If you crop a wide photo into a tall story, you may lose the product or the person. The other option is to extend the picture, which means software invents new background around it. That can work well, or it can leave a visible patch. This page helps you pick the right move for each shape.

I use Meta's own ad guidance, Adobe's and Canva's help pages, and Higgsfield's tool list, all checked on 3 November 2026. Where a number is not published, I say so. Prices and plans change, so confirm them on the vendor's page.

This page may contain affiliate links from which Blur Studio earns commissions.

The problem: one photo, three placements

An aspect ratio is the shape of the picture, written as width to height. A square is 1:1. A tall story is 9:16. A wide banner is often around 16:9 or wider. A single photo cannot be all three. You can make one, crop to another, and for the third, add new space around it.

Option 0: retake the photo, or recrop from a wider original

The cheapest, safest fix is a better source. Ask for the original file, because a photo sent through a chat app is often shrunk and may also be cropped. If you can still reshoot, take the picture wider than you need, with air above and below the subject. Then every shape is a crop, and nothing is invented. If you cannot reshoot, ask whether a wider frame from the same session exists.

Which shape goes where (Meta's own guidance)

Meta's business help page lists these ratios: 1.91:1, 16:9, 1:1, 4:5 and 9:16. Per that page:

Placement	Shape Meta recommends or supports
Stories, Status and Reels	9:16 recommended
Feed	1:1 and 4:5 supported. The page words it as 4:5 for single-image Facebook Feed ads and 1:1 for Instagram Feed.
In-stream video	16:9
Audience Network	9:16
Other horizontal shapes	1.91:1 and 16:9 are both listed as horizontal. Check which placement supports which.

One more point from Meta matters a lot. If you do not provide every ratio, the system fills the gaps itself with padding, media expansion, cropping and AI features. Meta tells advertisers to review the automatic crops. So your real job is to make your own 9:16 and 4:5 for the placements that matter, and to preview the rest. For the ad side of this work, see my page on resizing ads for every placement.

Crop first, extend second

Use this order:

1. List the shapes you need and where each one will run.
2. Crop for the ones that allow it. A 4:5 and a 1:1 are usually crops of a 4:5 or wider master.
3. Extend only the shape that needs more room, usually the 9:16.

A made-up number example: a 4:5 master at 1080 x 1350 pixels needs about 1080 x 1920 for a 9:16. Keeping the width, that is roughly 570 pixels to add, split between top and bottom. That is arithmetic only, but it shows the extended area is a band, not the whole picture.

Decision rule: extend when the new area is plain or repetitive, such as sky, a wall, a studio backdrop or soft blur. Do not extend when the new edge would show real surroundings that a customer could check, such as a shopfront, a room, a venue or a street.

Tools that extend an image

• Photoshop Generative Expand. Adobe's help page (updated 23 February 2026) says to drag the Crop tool outward to expand the canvas, choose Generative Expand in the contextual bar, pick one of Adobe's Firefly models (Firefly is Adobe's family of AI image models) such as Firefly Fill & Expand, leave the prompt blank if you like, and choose a variation. Check Adobe's page for current credit use and price.
• Canva Magic Expand. Per Canva's help page: Edit, then Tools, then Magic Expand, then choose a size or drag a handle. The expanded image downloads only as part of a design. Canva warns that images with faces, hands or transparent backgrounds may not expand well. That help page does not state which plan includes it, so check Canva's plans.
• Meta's own expansion. If a rough fill for Stories is enough and time is short, Meta's automatic fill can do it, but review the result.
• A designer. For a hero ad, a person can rebuild the background by hand and you can approve it.

Higgsfield: image Expand versus video Reframe

Higgsfield is an AI platform for making images and video. Its tools for this job are easy to mix up:

• Expand (outpaint). This extends an image beyond its edges. Higgsfield's tool list shows a FLUX.2 Pro Outpaint model with expansion set per side, up to 2048 pixels each side. It is also listed on the Apps page as "Expand image".
• Edit. This changes parts of an image using AI image models such as Nano Banana, Seedream or Flux Kontext, or a mask-based tool called Soul Inpaint (you paint over an area and describe what should replace it). Soul Inpaint lists 0.25 credits per image, and it is not on Free or Starter plans. Other edit costs are not published. To change one item in an approved image, see changing one item without redoing the whole image.
• Reframe. This changes the aspect ratio of a video, not a still image. It outputs 480p to 1080p.

So for a still photo, you would look for Expand, not Reframe. Higgsfield does not publish what an Expand costs, so read the credit cost on the button before you confirm. I explain the wider toolset in my guide to Higgsfield upscale, relight and edit tools and credits in how to spend Higgsfield credits wisely.

When does it fit? When you already work on Higgsfield, the photo is a generated or marketing image, and you want to stretch it to several shapes in one place. Free Higgsfield accounts watermark every output, so client files need a paid plan. See Higgsfield pricing and check higgsfield.ai/pricing for today's plans.

If you cannot extend: pad it, or protect the subject

Extending is not the only way to fill a new shape. Cloudinary's guide to aspect ratios (an image software company) advises adding background padding or borders to a non-standard ratio instead of stretching, because stretching distorts the picture. A plain colour or a blurred, enlarged copy of the photo behind the original is honest, because nothing is invented. A tool such as iResizer (which describes marking the objects you want to protect so they stay unscaled while the rest of the picture is squeezed) is another route when the background has no real detail to lose. Never stretch a photo to fit: people and products look wrong, and a buyer can notice.

The edge check list with pass/fail

Zoom in on the added area at phone size and at 100 percent. Look for:

• A seam where the original ends and the new part begins.
• Repeated textures that look like a copied pattern.
• Extra fingers, limbs or objects near the edge.
• Logos or text that are bent or changed near the border.
• Light that comes from a different direction, or shadows that stop.

Fail on any item a client could notice at phone size. If it fails twice, go back to Option 0 or crop instead. Also keep the logo and the call to action well inside the frame. Meta says to review crops so important content is not cut off. A figure for exact safe margins appeared in a search summary I could not verify, so I do not state one. Preview the placement instead.

Rights and truthfulness: when not to extend

• Real surroundings matter. A property, a shopfront or an event photo should not get invented extra space. A buyer may rely on what the picture shows.
• Client-supplied photos. Confirm the client allows AI editing, and the original licence allows edits.
• People. Do not use extension to change how a person appears. If the photo shows a real person, edits need their consent. Check local law and the platform's current rules.
• Marketplaces. Product listings often have strict image rules. Check the marketplace's current rules before you add background to a main image.

Export and naming for each size

Export each shape as its own file and name it so no one picks the wrong one. A simple pattern is: client, subject, ratio, version. For example: acme-shoe-9x16-v2.jpg, acme-shoe-4x5-v2.jpg and acme-shoe-1.91x1-v2.jpg. Save the master and a note of which versions used extension.

When to skip Higgsfield for this job

• You only need a crop. Any free editor does it.
• You already have Photoshop or Canva and only need an occasional extension. Their tools are inspectable, and Photoshop keeps layers on your machine.
• The photo shows a real place that must stay accurate.
• You need a published price before approving spend. Higgsfield shows Expand cost only at the button.

How I'd approach this

I would shoot or choose a master with space around the subject, make the 1:1 and 4:5 by cropping, and extend only the 9:16, a little at a time. I would run the edge check at phone size and keep the original beside every extended version. If I wanted to do it inside a paid Higgsfield account, I would try Expand on one picture and read the credit cost on the button before I did a whole campaign.

When an agency does this for many clients, the real time sink is the back and forth about which size is approved. Blur Studio builds the repeatable workflow, client approval portal or shared asset library around these tools, so every shape lives with its master and its approval.