Luma event cover image: square, 800 px minimum, text off the corners
Luma recommends a square 1:1 event cover at 800 by 800 px or more, static over GIF. How to generate one with Sume that survives rounded corners and small cards.

Luma's help page says event cover images should be square (1:1) for the best display across platforms, at least 800 by 800 pixels with no stated maximum, and that static images are recommended over GIFs. With Sume, request aspect_ratio: "1:1" and keep key content away from the corners and edges. One trap: a custom 800 by 800 size is below the 655,360-pixel minimum for GPT image models on Sume, so ask for 1024 by 1024 instead.
Luma's guidance is from Event Cover Images, read 2026-10-03. Sume's is from Image Router.
What does Luma's page recommend?
The page is short. These are its points:
- Cover images should be square (1:1), at least 800 by 800 pixels; there is no maximum, and larger images look better across sizes.
- Static images are recommended. GIFs are partially supported and not guaranteed, and Luma does not troubleshoot GIF issues.
- Minimize text overlay so the cover reads at several scales, and keep margins around the edges.
- Keep important elements out of the corners, because corner rounding can cut them off.
- Design for large pages and small preview cards alike.
Why not generate exactly 800 by 800?
Sume's Image Router docs state the custom-pixel rule for GPT image models: both edges multiples of 16, maximum edge 3840, aspect ratio at most 3:1, and 655,360 to 8,294,400 pixels overall. 800 by 800 is 640,000 pixels, below the floor. 1024 by 1024 is 1,048,576 and satisfies every rule, and it clears Luma's 800 pixel minimum with room to spare.
Or skip custom pixels. Send aspect_ratio: "1:1" and let the model use its own native square size, then confirm the returned width and height are at least 800 before you upload. Ratios are per model, so read the model's supported_parameters first.
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: luma-cover-001" \
-d '{
"model": "sume/auto",
"prompt": "Centered lantern on a dark table, wide empty margins, no text",
"aspect_ratio": "1:1",
"output_format": "png"
}'How do I keep a cover safe from corner rounding?
Luma says corners may be cut by rounding, so treat the outer margin as disposable. A practical rule is to put the subject inside the central area and leave empty background at each corner. Ask for that in the prompt: a centered subject, wide margins, and plain corners.
Words are the other risk. The page asks you to minimize text, and generated lettering is the thing most likely to come out wrong. Leave the cover text-free and let the event title live in Luma's own fields, where it is exact and editable.
What about an animated cover?
Luma's page recommends static images and says GIF support is partial. It does not document video covers, so this post does not describe any. If you want motion elsewhere, make the clip separately and keep the cover a still. Sume does not produce a GIF as a documented output of its image endpoint, so a still is the natural fit here.
A still from a finished event clip works too: video frames extracts 1 to 24 stills at times you name, as JPEG or PNG at the source size. Check that the frame is square-friendly first, since a 16:9 frame has to be cropped.
| Luma guidance | Sume setting | Check |
|---|---|---|
| Square 1:1 | aspect_ratio 1:1 | Model lists the ratio in supported_parameters. |
| At least 800 x 800 | image_size 1024x1024 on GPT models | 800x800 is under the 655,360-pixel floor. |
| Static over GIF | output_format png or jpeg | Sume docs list no GIF output for images. |
| Margins, no corner content | Prompt wording | Review the result at a small preview size. |
What should I check before upload?
Open the result's width and height from the job result (Jobs and results), view it at thumbnail size, and look at the corners. Sume does not upload to Luma; the file is yours to attach in Luma's editor.
Sources
Related posts
More in Use cases
- Mailchimp video merge tag: the thumbnail links out, so pick it well
Mailchimp video merge tags show a thumbnail that links to a URL, so the still is the ad. Choose the frame with Sume video frames: 1 to 24 stills, source size.
- Can Meta AI images @-mention my public Instagram account? The setting
Instagram's Stories effects from Meta AI let the feature use public accounts. What the page says, where the opt-out is, and what brands should do with Sume.
- Meta says phone-shot Reels beat polished video: what Sume can do
Meta's holiday guide says phone-shot Reels outperform polished brand videos. Keep real phone footage, then use Sume to trim, caption and add generated b-roll.
- Meta's Sept 3 advertiser results: test the claims on your own ads
Meta published five advertiser results for its AI ad tools, from 64% lower cost per purchase to a 4.4% sales lift. How to read them and run a test of your own.
Written by Sume