Images on X: the complete guide

By Sami KerrPublished 18 min read
Featured image for Images on X: the complete guide. Three picture cards fan out on a black stage: a landscape photo with a sun over hills, a bar chart, and a quote card with a large opening quote mark, with a small ALT badge on the front card. Beside them the title Images on X: the complete guide and the label Images, video and design.

The short answer

An image earns its place on X when it does a job words would do worse: show a thing, prove a claim or compare two things. Pick the kind for the job, keep text big and high in contrast for a phone, describe every image, and post only pictures you have the right to use.

Key takeaways

  • Cover the image. If the post reads the same without it, the image is decoration, so leave it out or find a better one.
  • X sets no pixel size for post photos. A single photo with a shape between 2:1 and 3:4 displays in full, and a photo can be up to 5 MB.
  • Text in an image needs a contrast ratio of at least 4.5:1 for normal text, and it must still be readable when the image shrinks to phone width.
  • Describe every image. An image description can be up to 1,000 characters, and any words inside the image belong in it.
  • Giving credit is not permission. Use your own images, licensed images or images the owner said yes to, and say so when an image is AI made.

You have a post to write and a picture you could attach to it. Every one of your X images starts with the same question: does this picture earn its place, and will it still do its job on a phone, for a reader who may never see it at all?

This guide follows the path an image takes. First whether to use one, then which kind, then sizes, text, contrast, the file, the description and the rights. Where a sibling guide goes deeper, we give the short version and link to it. If an upload comes out soft, see why X images look blurry and how to fix it.

When does an image help an X post

An image helps when it does a job words would do worse. It shows a thing, proves a claim, compares two things or squeezes a dense list into one glance. If the post reads the same with the picture covered, the picture was decoration, and decoration costs your reader a moment for nothing.

The cover test for an image in a post Two post cards side by side. On the left, a post reading Our new office is open has a landscape picture. A blind drops over the picture and the label reads Nothing lost, so it is decoration, in rose. On the right, a post reading Reply time, six weeks has a line chart that falls from 48 to 21. A blind drops over the chart and the label reads The data is gone, so it has a job, in mint. Our new office is open. Nothing lost so it is decoration Reply time, six weeks: 48 21 The data is gone so it has a job
The cover testCover each picture. The left post loses nothing, so its image is decoration. The right post loses its number, so its image has a job.

Four jobs pass the cover test again and again.

Job 1

Show

A product in use, a place, a result on a screen. The reader sees what words would need a paragraph to describe.

Job 2

Prove

A screenshot of the message, the receipt or the dashboard you mention. It turns "trust me" into "look".

Job 3

Compare

Before and after, or two options side by side. The eye compares shapes faster than it reads two sentences.

Job 4

Compress

A short table or list that would take three posts. One clean image is easier to take in.

Images do not use up your 280 characters. In X's official apps, attached media counts as 0 characters, as we explain in how X counts characters. That makes a picture a good home for something dense. It does not make it a good place to hide the point, because a screen reader cannot read words inside a picture unless you describe them.

Which kind of image should you use on X

Six kinds cover almost everything you will post: photos, screenshots, charts, text images, GIFs and memes. Pick the kind by its job, then check the one risk each kind carries. For charts and data, see how to design an infographic for X. Three more image guides: news cards for breaking stories, GIFs that load fast and colour contrast people can read on phones.

A contact sheet titled Six kinds, six jobs. Frame 01 Photo: show the real thing. Frame 02 Screenshot: prove what happened. Frame 03 Chart: compare one thing. Frame 04 Text image: make a line stand out. Frame 05 GIF: show one short step. Frame 06 Meme: share a joke or mood.
One job per kind. If your picture does not fit one of these jobs, it is probably decoration.
KindBest forWatch forFile we export
PhotoShowing a real thing, place or personSubject near the edge, a file over 5 MBJPEG
ScreenshotProving what something said or didPrivate details, tiny textPNG
ChartOne comparison or one trendMeaning carried by colour alone, tiny labelsPNG
Text imageA line, number or announcement that must stand outToo many words, low contrast, no descriptionPNG
GIFOne short moving stepBig files, flashing, rough colourGIF
MemeA shared joke or reactionRights to the picture, no descriptionPNG or JPEG

The next sections take each kind in turn, with the check we run before it goes out.

Photos and screenshots: what to check

Use your own photos wherever you can, because they are the one kind you never need permission for. Put the subject near the middle, give it room to breathe and keep words away from the edges. We do not know exactly how X crops each picture in a post with several, so a centred subject is cheap insurance.

A screenshot is proof, so it has to be readable. Crop to the one panel that proves the point, enlarge the line that matters and put a box around it. A full desktop with fifteen tabs open proves nothing.

Cropped to the proof

One panel, the key line boxed, the customer's email covered with a solid block. The reader sees the point in two seconds.

The whole screen

A full desktop with an inbox, a name and a phone number visible. The point is lost and the private detail is out.

Privacy comes next. X's rules say you may not publish other people's private information without their authorisation, or share private media of individuals without their consent. One frame of an inbox, a direct message or a customer's order can break that. We cover private details with a solid block, flatten the file and export a fresh PNG.

How to make a chart readable on X

A chart works on X when it makes one comparison obvious on a phone. One chart, one claim, and the title states the claim.

A line chart with example data, annotated in four steps. The title reads Reply time halved after the rota change. The y axis is Median reply time, minutes; the x axis runs from Week 1 to Week 6. The line starts at 48 minutes, stays near 50 for three weeks, then falls to 21 minutes. A marker at week 4 reads Rota change. Annotations: 1 Title says the finding. 2 Label the line, skip the legend. 3 Mark the event. 4 Name the source. The footer reads Source: support desk log, example data. A tag reads Example.
A made up chart, marked up. Every number in it is example data.

Five habits make the difference.

  • Title the finding. "Reply time halved after the rota change" tells the reader what to see. "Reply time by week" makes them work it out.
  • Label the line itself. A legend sends the eye back and forth. Put the name or the value next to the line.
  • Do not lean on colour. WCAG says colour should not be the only way to tell things apart, and about 1 in 12 men have colour vision deficiency, according to the US National Eye Institute. Add a label or a different line style.
  • Keep chart marks clear of the background. WCAG asks for at least 3:1 contrast for graphical objects needed to understand the content, such as lines and pie slices.
  • Name the source. One small line under the chart.

A chart also needs a description, because a screen reader gets nothing from the picture alone. The W3C advises a short description that names the type of chart and a longer one with the values and trends. On X you have one box of up to 1,000 characters, so put the type and finding first and the key values after. A worked example is in the alt text section below.

Text images and quote cards

Use a text image when one line should stand out and travel on its own: a quote, a single number, an announcement. Skip it when the words are the post. The reason is practical. A screen reader cannot read words inside a picture, and WCAG asks for real text instead of images of text where the technology allows it.

So we keep two habits:

  • Keep it short. About 20 words, which is the guide inside the free quote card maker.
  • Repeat the full text in the image description, so nobody is left out.

Our guide to making quote cards for X covers choosing a quote, attribution and contrast in depth. The maker builds the card in your browser at a shape X shows in full.

GIFs and memes: size, flashing and descriptions

Use a GIF for one short step that needs motion, such as a click path or a before and after flip. Three limits are worth knowing:

  • File size. In the app, an animated GIF can be up to 5 MB on mobile and up to 15 MB on web.
  • Colour. A GIF holds at most 256 colours, so a photographic scene often turns grainy.
  • Flashing. WCAG 2.3.1 says nothing should flash more than three times in any one second, to protect people with photosensitive seizure disorders. It says short clips that may loop, such as GIF animations, are judged while looping.

X does not support animated GIFs for a profile photo or header. Your smallest image matters too; see how to choose a profile picture people recognise.

A meme is someone's picture with new words, which gives it two checks:

  • Do you have the right to the picture? The rights section below applies in full.
  • Can a person who cannot see it get the joke? Describe both the picture and the text.

Few people do. A Carnegie Mellon University study from 2019 looked at 9 million posts and found about 1 million with images. Only 0.1% of those images had a description. A meme that carries one still stands out. Whether a meme fits your account is a voice decision, and brand voice for X is the place to settle it.

Image sizes and crops for X, in brief

X sets no pixel size for photos in a post. It gives one rule about shape: a single photo with a shape between 2:1 and 3:4 displays in full. The header and the profile photo have recommended sizes.

2:1 to 3:4

Shapes a single photo shows in full

X Help Center, posting photos

5 MB

Largest photo in a post

X Help Center and X API documentation

1500 × 500

Recommended header, in pixels

X Help Center, customize your profile

400 × 400

Recommended profile photo, in pixels

X Help Center, uploading a profile photo

The full table, with the maths for the header's safe band, lives in X image sizes and dimensions. The free image size guide and cropper crops any image to the right size in your browser and shades the crop zones. Before you post someone else's picture, read our guide to copyright on X.

Three crops over one picture, and what survives One landscape picture with a round amber subject in the middle and four words near its edges: Launch day at the top, Free shipping at the bottom, New on the left and Today on the right. Three example crops, wide, square and tall, are drawn as dashed outlines. One after another, each crop dims everything outside it. The wide crop cuts the top and bottom words, the square and tall crops cut the left and right words, and the subject in the centre is never cut. The crops are examples, not X rules. Launch day Free shipping New Today wide square tall
Design for the middleThree example crops slide over one picture. Words at the edge are cut by some of them. The subject in the centre survives all three. The crops are examples, not X's rules.

We start from 1600 by 900 for screenshots and charts, 1080 by 1080 for cards and logos, and 1080 by 1350 for portraits. All three sit inside the 2:1 to 3:4 range. They are our choices, not X's numbers.

How many images can one X post hold

A post can hold up to 4 photos. X does not publish how each picture is cropped when there are several, so plan every image to survive a crop: subject in the middle, words away from the edges.

A strip of four numbered frames with the heading Four images, one story. Frame 1 The result. Frame 2 The setup. Frame 3 The detail. Frame 4 The proof. Under the strip: The post text carries the point. Each frame adds one thing.
Give each image one job. If two frames do the same job, drop one.

Four rules we follow:

  • One job per image. A second frame should add something the first does not.
  • Make the first image work alone. We cannot promise which frame a reader opens first.
  • Keep the point in the words. The post text carries the claim, and the pictures carry the evidence.
  • Describe every image. X lets you add a description to each image in a post.
Can you mix photos, a GIF and a video in one post?

It depends on where you read. X's Help Center says a post can hold up to 4 media items in total, for example 2 photos, 1 GIF and 1 video. X's API documentation describes a post with up to 4 photos, or 1 animated GIF, or 1 video. The two pages describe different users, and we could not open the Help Center page to settle it, so we plan a post around one kind of media unless we have tried a mix in our own app.

How big should text be on a phone

Text in an image should be large enough that it is still readable when the whole image shrinks to a phone's width. Design at a wide size such as 1600 pixels, then do the maths. A phone about 360 points wide shows your image at 360 divided by 1600, which is 0.225 of its design size.

Text sizes on a 1600 pixel canvas and on a phone The scale factor is 360 divided by 1600, which is 0.225. Three text sizes are drawn as bars to scale. A headline of 96 pixels shrinks to 22 points on a 360 point wide phone. A main line of 64 pixels shrinks to 14 points. A credit line of 48 pixels shrinks to 11 points. The phone width is an example; phones vary. 360 / 1600 = 0.225 Headline 96 px 22 pt Main line 64 px 14 pt Credit line 48 px 11 pt
From the canvas to a phoneThree text sizes drawn to scale, each shrunk to 22.5 per cent. The 96 pixel headline lands near 22 points, 64 near 14 and 48 near 11. The phone width is an example; phones vary.

That gives a simple table for a 1600 pixel canvas.

Text size on the canvasAbout this on a 360 point phoneWe use it for
96 px22 ptThe headline or the big number
64 px14 ptThe main line of a card
48 px11 ptA credit line, nothing the reader must have

Fewer words make bigger type possible. A card with 20 words can carry 64 pixel text. A card with 60 cannot, so cut words before you shrink type.

Contrast and colour: images that work in light and dark

WCAG level AA asks for a contrast ratio of at least 4.5:1 for normal text and at least 3:1 for large text. Large means 18 point, or 14 point if bold. A ratio runs from 1:1 (no contrast) to 21:1 (white on black).

We worked out the ratios below with the WCAG formula, for pairs that people really use.

Text on backgroundRatioNormal text (4.5:1)Large text (3:1)
White on black21.0PassPass
Off white f0f0f0 on black18.4PassPass
Dark grey 4b5563 on white7.56PassPass
Grey 767676 on white4.54Pass, justPass
Grey 777777 on white4.48Fail, justPass
Light grey 9ca3af on white2.54FailFail
Light grey b8bbbd on white1.93FailFail

One colour can work in one mode and vanish in the other. Amber ffc15e on black is 13.04:1, and the same amber on white is 1.61:1. Because people read X in light and dark, check both. To make every image look like it comes from you, see how to keep a consistent visual style on X.

Which file type and file size for X images

Export PNG for text, lines and flat colour, and JPEG for photographs. A photo in a post can be up to 5 MB, and the app accepts GIF, JPEG and PNG. The API documentation also lists WEBP, so the two do not match, and we leave WEBP out of anything that has to work everywhere.

PNGJPEG
Lossless: edges around letters stay sharpLossy: small type gets soft edges and coloured fringes
The format to prefer for screenshots, diagrams and logosA good choice for photographs
Supports transparencyNo transparency
The edge of a letter saved as PNG and as JPEG Two magnified pixel grids of the same letter edge, light on dark. The PNG grid has a hard edge with two clean colours. The JPEG grid has off colour pixels along the edge, tinted rose and blue, which is why small type looks fuzzy when it is saved as JPEG. The pixels are drawn for illustration. PNG JPEG 2 clean colours off colour pixels
The edge of a letter, magnifiedThe same letter edge saved two ways. PNG keeps two clean colours. JPEG leaves off colour pixels around the edge, which is why small type looks fuzzy.

Three more habits help. Export once, at the final size, from your original. Start from a picture at least as large as you need, because enlarging a small one makes it soft. And keep the file under 5 MB for a photo.

Other images have their own rules. The image for a link preview card must be under 5 MB, and ads follow separate creative specs. Our table of limits is in X image sizes and dimensions.

Alt text for every image on X

X calls alt text an image description. It can be up to 1,000 characters, and X's advice is to include any text that appears in the image. You add one after attaching the image by selecting +ALT, typing and selecting Done. An ALT badge then shows on the image.

  1. Attach the image

    Add it to your post as usual.

  2. Select +ALT

    A text box opens on the image.

  3. Say what it shows and why

    Start with the point. Write out any words in the image.

Write it before you post. We found no X page about editing a description once a post is live, so treat it as fixed. You can also switch on a reminder in Accessibility settings that warns you when an image has no description.

Pilar Moss (invented example)@pilar_exampleExample

We moved support onto a rota in week 4. Median reply time went from about 50 minutes to about 20. The chart shows the six weeks.

Invented for this guide, with example data

Type and finding first

"Line chart of median reply time in minutes over six weeks, example data. It sits near 50 for three weeks, then drops to about 20 after the rota change in week 4."

A label, not a description

"Chart." Or "Image of a graph." A reader learns nothing the post did not already say.

Our full guide to writing alt text for images on X covers length, tone, screenshots, memes and the mistakes that are easy to make.

Can you use that image? Rights, permission and privacy

Post only images you made, images you hold a licence for, or images the owner has said yes to. Giving credit is not permission. The US Copyright Office says you cannot claim copyright to another's work, no matter how much you change it, without the owner's consent.

It also says there are no legal rules permitting a set number of words, notes or a percentage of a work. Limited uses such as quotes for commentary, criticism or news reporting may be allowed, and the Office recommends permission "in cases of doubt". That is US law, and this is general information, not legal advice.

Safest

Your own

Photos you took, charts you built, cards you designed. Check that any font, icon or stock element inside them is licensed too.

Read the terms

Licensed

A stock site or a commissioned photographer. Licences differ by site and by use, so read what yours allows before you post.

Four conditions

Creative Commons

Attribution means credit the creator. NonCommercial limits you to noncommercial use. NoDerivatives bars edits. ShareAlike asks you to share adaptations on the same terms.

Get it in writing

With permission

Ask the owner, and keep the reply. A written yes beats a memory when anyone asks later.

X can act on reports. Its copyright policy covers a copyrighted image used as a profile or header photo or uploaded through media hosting services. If X removes material, it notifies you and explains how to send a counter notice.

Other rules touch images too. X may treat a trademark used to mislead people about your connection to a brand as a violation. A sponsored image needs the Paid Partnership disclosure. X allows consensual adult content when the post or account is marked sensitive. Our guide to X's rules for brands and creators explains each.

Synthetic and manipulated images on X

X's rule now sits in its Authenticity policy: you may not share inauthentic media, including manipulated or out of context media, that may cause widespread confusion on public issues, affect public safety or cause serious harm. That covers media shared as real but significantly and deceptively altered or fabricated, and media shared with false context. X's enforcement page says it may label posts with misleading information.

That is a high bar, and our own bar is lower. We label an image as AI made whenever it could pass as a photograph of a real event, person or product. We never make an image of a real person doing or saying something they did not.

Fine, and we label itNot for us
An AI illustration of a concept, marked as AI madeAn AI photo of a real event that never happened
Cropping and colour correcting your own photoEditing a screenshot to change what a person said
A composite, with the edit mentioned in the postA fake quote card in someone's name

Labels have a technical side. X's developer documentation for creating a post includes a field to disclose that a post contains AI generated media. We found no Help Center page that explains how an app user sets the same disclosure, so check the composer in your own app.

The Kairo Post AI Agent page shows how its pictures and designs are made and kept, with every version, in the Library. Whatever you use, keep that record yourself: who made the image, with what, and when.

Video on X, in brief

Video follows different rules from images. Without X Premium you can upload a video up to 140 seconds long and up to 512 MB. Subscribers can upload far longer ones. MP4 is the file type named across X's pages.

ImageVideo
Main limitUp to 5 MB per photoUp to 140 seconds and 512 MB without Premium
Words for people who cannot see itImage description, 1,000 charactersCaptions with an SRT file
File we exportPNG or JPEGMP4

Our guide to video on X has the limits, formats, captions and thumbnails.

A checklist to run before any image goes out

This is the order we work in. It takes a few minutes, and it catches almost everything in this guide.

Before the image goes out

  • The image passes the cover test: the post loses something without it.
  • The kind fits the job: photo, screenshot, chart, text image, GIF or meme.
  • The shape suits the post, 2:1 to 3:4 for a single photo, and the subject sits in the middle.
  • Text is large enough at 360 pixels wide, and the contrast is at least 4.5:1.
  • A chart has a title that states the finding, direct labels and a source line.
  • Text and line art are PNG, photographs are JPEG, and the file is under 5 MB.
  • The image has a description, with any words in it repeated.
  • You made it, hold a licence or have written permission.
  • No private details show, and no person appears without their consent.
  • If it is AI made and could pass as real, the post says so.
  • You looked at the post in the preview, light and dark.

If you post many images, doing this by hand gets slow. The Kairo Post AI Agent page describes designs made in your brand kit, AI pictures that follow one saved image style per account and alt text you can draft in one click and edit before publishing.

For your next post, start with the cover test. If the image passes, pick the kind, export it at the right size and describe it before you post.

In the product

How Kairo Post does this

Everything above works by hand. Kairo Post does the same work for you, with your approval before anything goes out.

Questions people ask

Do images get more reach on X?
We found no X page that says so, and we do not make that claim. X's help pages describe how images are posted and shown, not how they rank. The fair test is your own: compare similar posts with and without a picture in your analytics, and judge by what your readers do.
How many images can I add to one X post?
You can add up to 4 photos. X's Help Center describes up to 4 media items in total, for example 2 photos, 1 GIF and 1 video, while the API documentation describes 4 photos, or 1 GIF, or 1 video. If your app does not accept a mix, that is why.
What size should an image be for an X post?
X sets no pixel size for photos in a post. It says a single photo with a shape between 2:1 and 3:4 displays in full. We design at 1600 by 900 for landscape, 1080 by 1080 for square and 1080 by 1350 for portrait, which all sit inside that range.
Should I use PNG or JPEG for images on X?
Use PNG for anything with text, lines or flat colour, such as screenshots, charts, logos and cards. Use JPEG for photographs, where the smaller file is worth the loss. The app accepts GIF, JPEG and PNG, so we avoid WEBP, which only the API documentation lists.
Can I post an image I found online if I credit the owner?
Not safely. The US Copyright Office says giving credit does not authorise a use, and that there is no safe number of words or percentage you can take. Use your own image, a licensed one, or ask the owner. This is general information, not legal advice.
Do I have to label AI generated images on X?
We found no X page that makes a label compulsory for every AI image. X's rule is about synthetic or manipulated media that may deceive people and lead to harm. Our own practice is to say so whenever an AI image could pass as a real photograph.
Do images count towards the 280 characters?
No. X's developer documentation says attached media counts as 0 characters when you post from the official clients. An image description has its own limit of 1,000 characters, separate from the post.

Sources

Every rule, limit and figure in this post was checked against these on the date shown.

  1. How to post photos or GIFs on X, X Help Center

    help.x.com · checked

  2. How to Post, X Help Center

    help.x.com · checked

  3. How to add image descriptions, X Help Center

    help.x.com · checked

  4. Copyright policy, X Help Center

    help.x.com · checked

  5. Authenticity, X Help Center

    help.x.com · checked

  6. Private information policy, X Help Center

    help.x.com · checked

  7. Counting Characters, X developer documentation

    docs.x.com · checked

  8. Create Post, X developer documentation

    docs.x.com · checked

  9. Complex Images, W3C Web Accessibility Initiative

    w3.org · checked

  10. Color blindness, National Eye Institute

    nei.nih.gov · checked

  11. Image file type and format guide, MDN Web Docs

    developer.mozilla.org · checked

  12. Copyright FAQ on fair use, U.S. Copyright Office

    copyright.gov · checked

  13. About CC licenses, Creative Commons

    creativecommons.org · checked

Written by

Sami Kerr, founder of Kairo Post

Keep reading

All posts on this topic

Post at the right moment.
Starting today.

Connect an account, describe your voice and let Kairo Post do the legwork.