LinkedIn Message ads specs: 300x250 banner, 60-character subject
LinkedIn Message ads take a 300x250 JPG, GIF or PNG up to 2 MB, a 60-character subject and 1,500 characters of text. Sume's image size floor and the fix.

A LinkedIn Message ad uses a banner creative of up to 300 x 250 pixels as JPG, non-animated GIF or PNG at a maximum of 2 MB, a subject line of up to 60 characters, and message text of up to 1,500 characters. The banner shows on desktop only. Those limits come from LinkedIn's Message ads advertising specifications page, read on 2026-10-03.
The banner is tiny, and the text carries the ad. That changes where AI helps: it is better spent on the copy and one clean banner than on a large hero image.
Every limit on the page
The page notes that a long subject line may be truncated and recommends keeping it concise. Count macros in the 60 characters: a personalization token that expands to a long first name can push a line over.
| Element | Limit |
|---|---|
| Banner creative | 300 x 250 pixels maximum, desktop only |
| Banner file type | JPG, non-animated GIF or PNG |
| Banner file size | 2 MB maximum |
| Subject line | 60 characters, including spaces, punctuation and personalization macros |
| Message text | 1,500 characters; up to 10 emojis; bold, italics and bullets supported |
| Call-to-action button | 20 characters; one primary button |
| Not Interested button | Required, added automatically |
| Hyperlinked text | 70 characters |
| Clickable links | 3 maximum |
| Custom footer | 20,000 characters |
| Destination URL | 1,024 characters, must start with http:// or https:// |
The Sume image size floor
Sume cannot render a 300 x 250 file directly on the GPT image model. The Image API says a custom image_size for openai/gpt-image-2.5 needs both edges to be multiples of 16 and between 655,360 and 8,294,400 pixels in total, and 300 x 250 is only 75,000 pixels.
Ask for a larger image at the same shape and scale it down yourself. 300 x 250 is 6:5, and 1440x1200 is also 6:5, with both edges divisible by 16 and 1,728,000 pixels, so it passes. Export that as PNG or JPEG, then resize to 300 x 250 in an image editor and save under 2 MB. Sume has no image resize step in its docs, so the downscale is outside Sume. Check the result at 300 x 250 before you upload, because small text that read well at 1440 x 1200 may not.
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: msg-banner-001" \
-d '{
"model": "openai/gpt-image-2.5",
"prompt": "Clean product banner, one object centered on a flat pastel background, no text, wide margins",
"image_size": "1440x1200",
"quality": "high",
"output_format": "png"
}'Keep words out of the banner
At 300 x 250 an image model's lettering will not survive, so ask for no text in the picture and put your words in the subject line, the message and the 20-character button. The image is a background for the sender's identity and one product or idea. Review the generated picture for stray lettering before the downscale; AI images can add text you did not ask for.
Then write the copy to the limits. Draft the subject at 60 characters or fewer, the message under 1,500, and a button label of 20 or fewer. For the same reason as the banner, the Not Interested button is automatic, so do not spend a character on it.
What to check before you send
- The banner is 300 x 250 or smaller and under 2 MB, in JPG, PNG or a still GIF.
- The subject is 60 characters or fewer with macros expanded.
- There are no more than three clickable links and ten emojis.
- The destination URL starts with https:// and is under 1,024 characters.
- You have looked at the banner at its real size on a desktop screen, since mobile does not show it.
Sources
Related posts
More in Use cases
- LinkedIn Thought Leader ads: which posts qualify and which don't
Thought Leader ads can sponsor a member's public text, image, video, article or Live event post, not documents, polls or multi-image posts. What to prepare.
- Lipstick shade preview with image edit: what a render cannot promise
A shade preview from gpt-image-2.5 on Sume is a picture, not a colour match. How to build it with a mask, and what to tell shoppers.
- Live AI avatar for streaming: pre-rendered clips instead
Streamers asking for a live AI avatar get conversation APIs. Sume cannot go live, but renders avatar intros, stingers and ad reads. Tavus greenscreen vs Sume.
- LIVE GMV Max wants 50-70 videos in queue: how to supply them
TikTok says add at least 50-70 videos in the In queue status for stable LIVE GMV Max exploration. One Sume bulk queue can cover it; here is the order of work.
Written by Sume