Articles / LinkedIn Image Post Dimensions: The Window, Not the Number

linkedin

LinkedIn Image Post Dimensions: The Window, Not the Number

Finn ·

For a photo you upload to a LinkedIn post, export at 1080 pixels wide or more and keep the width to height ratio between 3:1 and 4:5. LinkedIn's help page sets the minimum at 552 by 276 pixels and the upload limit at 5 MB. Outside that range, the image is centered and cropped to fit. The 1200 by 627 you have read everywhere is a real number from a different spec.

The three sizes worth exporting

Pick one of these and you are inside the documented window every time.

  • 1080 by 1350, ratio 4:5. The tallest shape the documented frame allows, so the most column space one photo can occupy.
  • 1200 by 1200, ratio 1:1. The square that survives every layout, and the size LinkedIn recommends for a square single image ad.
  • 1200 by 628, ratio 1.91:1. The landscape option, the same shape as the link preview card.

The window matters more than any of those three numbers, because the window is what crops you. LinkedIn's share photos or videos help page, checked 11 September 2026, states it plainly: "The photo's aspect ratio can range from 3:1 to 4:5 (width:height). If the ratio is larger, the image will be centered and cropped to fit the max ratio."

So divide width by height before you upload. Above 3.0 or below 0.8, you are outside the range and LinkedIn decides the framing instead of you.

Here is the case that catches most people. A screenshot straight off an iPhone is 1179 by 2556 pixels. Divide and you get 0.46, far outside the range. The same page puts the maximum height of the image container at 4:5, so at 1179 pixels wide the tallest available frame is 1179 divided by 0.8, which is 1474 pixels. That last figure is arithmetic on a documented frame rather than a number LinkedIn publishes, and it puts roughly 1082 pixels of the screenshot outside the feed, split top and bottom: the app bar, the total at the bottom of the invoice, gone, and you find out after publishing.

The fix is a canvas rather than an export setting. Drop the screenshot onto a 1080 by 1350 background and scale it to fit inside. Then check the composer preview, because the same page is equally plain that "Uploaded photos can't be resized after they have been posted."

If fitting everything into one image makes the text unreadable, compare LinkedIn image, document and video post formats before shrinking it further. The worked example shows how the same idea changes across containers.

Why LinkedIn publishes three different numbers

Because three help pages answer three different questions, and only one of them is about a file you upload.

A photo in a post follows the page above: 552 by 276 minimum, 1080 wide recommended, 3:1 to 4:5, 5 MB per upload.

A Page post that shares a URL follows the Pages image specifications, which say "Use a 1.91:1 ratio (1200 x 627 pixels)", PNG or JPEG, 3 MB maximum. That image is the link preview card, built from the tags on the destination page rather than from a file you attach, so you fix it in your HTML and not in the composer. LinkedIn Post Inspector: What It Fixes and What It Cannot walks that path.

A single image ad follows the ads specifications: 1200 by 628 horizontal, 1200 by 1200 square, 720 by 900 vertical, 5 MB, and anything narrower than 401 pixels displays as a thumbnail.

That is where the famous 1200 by 627 comes from. It is not wrong for a photo post, since 1.91 sits comfortably inside the 3:1 to 4:5 range. It answers a question about link cards, so it says nothing about height, and height is the edge of the window most people actually hit.

A fourth page, media file types supported on LinkedIn, lists GIF, HEIF/HEIC, JPEG, PNG and WEBP, that last one on Android only, then notes a 100 MB ceiling and a 36 megapixel limit covering images, documents, video and audio together. The photo page says 5 MB for an upload, so export under 5 MB and the contradiction never reaches you.

A size guide you saved is a timestamp, not a spec: LinkedIn moved the tall edge of its photo frame from 2:3 to 4:5 in early 2024.

What that same page said in early 2024

The size guide you saved two years ago is not sloppy. The spec moved underneath it.

The archived snapshot of that page from 20 February 2024 reads: "The photo frame ratio ranges from 3:1 to 2:3 (width/height)". The snapshot from 30 March 2024 reads what the live page reads today, 3:1 to 4:5, and adds the 1080 pixel recommendation that was not there before.

As decimals, 2:3 is 0.667 and 4:5 is 0.8. The tall edge of the frame came down, so LinkedIn renders less height than it documented in February 2024. A 1000 by 1500 portrait was inside the frame then. Today a 4:5 frame at 1000 pixels wide stops at 1250 pixels of height, so 250 pixels of that export sit outside it. Nothing about your file changed. The frame did.

The Page cover image makes the point faster: 1128 by 191 pixels on the specifications page in January 2024 and still in August 2025, 4200 by 700 in the February 2026 snapshot, 1512 by 256 today. Three documented numbers for one slot in about a year.

How far back the 2:3 frame goes is a separate question, since the archived record for that page begins in early 2024. Treat older numbers as unverified rather than wrong. A star rating ages the same way, which is the argument in Base44 Reviews: The Star Rating Is a Timestamp, Not a Verdict: the number was true when it was written, and nobody sends you a correction.

What survives the next change

Export wider than you need. 1080 pixels is the floor LinkedIn recommends, and a file sitting on that floor has nowhere to go when the frame moves. A 2160 pixel wide export at the same ratio costs you nothing and still fits under 5 MB as a JPEG.

Design inside the window rather than on its edge. A template sitting at exactly 4:5 is one frame change away from being cropped, and square has stayed inside the range on every one of these pages.

Read the specification page for the surface you publish on, and note the date you read it. Then look at the preview before you confirm. That is the easy step to skip when posts go out on a schedule, which is part of the case in Scheduled Post on LinkedIn: Be There When It Goes Live. If a tool publishes for you, the crop rule applies to the file it uploads rather than the file you designed, so check what it sends, ReadyToPost included.

Open the last image you posted, divide its width by its height, and see which side of 0.8 it landed on. That number decided what your readers saw.

FAQ

How many images can I put in one LinkedIn post? Up to 20 in a multi-photo post. The layout follows the first image's height and orientation, the container still maxes out at 4:5, and images in the bottom row can be cropped to fit.

Can I fix an image after the post is published? No. The help page states that uploaded photos can't be resized and can't be edited once posted, so a wrong crop means a new post rather than a correction.

Did this article help?

Get the best articles, carefully selected to save you time.

Read next

SaaS marketing automation is the set of messages your product sends because of what a user did inside it: signed up, reached the first useful result, went quiet, ran out of trial. The hard part is not picking a tool from a list of thirty. It is naming the one event that means a user got value, then hanging every trigger and every exit condition off that name.

A bad news email subject line should identify the affected service, order or request. For an operational change, include the change and date: "Your Pro plan rises to $29/month at your 12 November renewal". For a sensitive personal decision, a neutral subject naming the request can be more appropriate. Put the explanation and next steps in the body.

Featured

A pivot is often just the polite word we use with investors when the first company is dead and we have decided to build another one. And that is fine. Not because failure is noble, but because luck needs exposure: every market you enter, every product you ship and every channel you test is one more surface where something unexpected can land.

Build articles

A self-hosted AI agent is an agent whose runtime you install on hardware you control. In most setups that covers only one of three layers: the runtime is yours, the model is still a cloud API, and the tools reach wherever you point them. Keeping the model local is a separate decision, and it costs memory, roughly 16 GB for the smallest capable models, plus some tool-calling reliability.

Agentic means having agency: able to act toward a goal on your own behalf, or on someone else's. Applied to software, a system is agentic to the degree that a model, not code you wrote, chooses the next step: it takes a goal, picks which tool to call, and loops until it decides the job is done. It is a spectrum rather than a label, and the word is decades older than AI.

Projects

Brands

The essentials, by email.

What works, what does not, what I would do differently. Sent when I have something useful to say.