Square 1:1 ad video: which Sume models accept it, and 4:5
Seedance, Kling, Wan and MiniMax accept 1:1 on Sume; Gemini Omni takes only 16:9 and 9:16. No video model lists 4:5, so Feed needs a crop. Table and steps.

For a square 1:1 ad video on Sume, use Seedance 2.x or 2.5, Kling 3, Wan 3.0 or MiniMax H3 and H3 Max, all of which list 1:1 in their aspect ratios. Gemini Omni Flash 1.1, which is also the default for sume/auto, lists only 16:9 and 9:16, so a request for 1:1 will not work there. No video model in the Sume catalog lists 4:5, which is the ratio Meta recommends for Facebook Feed video, so a 4:5 ad needs a conform step after generation.
Aspect ratios per model
The values below come from the capability records in the Sume repository. The public /v1/videos descriptor shows the subset a caller may send as aspect_ratio, so confirm with GET /v1/videos/models before you code against a list.
| Model id | Aspect ratios |
|---|---|
| seedance-2, 2-mini, 2-fast, 2.5 | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 (plus auto and adaptive in the router) |
| kling-3 | 16:9, 9:16, 1:1 |
| wan-3.0 | 16:9, 4:3, 1:1, 3:4, 9:16 (plus auto and adaptive) |
| minimax-h3, minimax-h3-max | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 (plus adaptive) |
| gemini-omni-flash-1.1 | 16:9, 9:16 |
When 1:1 is the right call
A square works in many feed positions and on desktop, and it is the ratio a lot of existing creative libraries already have. If your ad set includes a square placement, generate the square natively on a model that lists 1:1 rather than cropping a 16:9 clip, because a native square composes the subject for the frame. A center crop of a wide clip often cuts a product off or leaves it small.
Kling 3 is the narrowest of the models that offer 1:1. It lists three ratios and resolutions of 720p and 1080p, so use it when those are enough. Seedance gives the widest range if you need to compare squares with 4:3 and 3:4 in the same test.
What 4:5 needs
Meta's Facebook Feed page, read on 2026-10-07, lists 4:5 at 1440 by 1800 as the ratio and recommended resolution. Since no video model lists 4:5, generate at 3:4 or 9:16 and conform. Timeline 1.0 accepts output width and height as even integers from 256 to 2160 and defaults to cover, so a 3:4 source rendered at 1440 by 1800 fills the frame and trims a small band from the top and bottom, as 3:4 is a little taller than 4:5.
If the product sits near the edge of the frame, prefer contain or blur fit, which keep the whole image and fill the rest of the frame.
curl -sS -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: square-mug-v1" \
-d '{"model":"kling-3","prompt":"Mug on a wooden desk, slow push-in","duration":5,"aspect_ratio":"1:1","resolution":"720p"}'A short decision list
Choose by the placement list, not by habit. Write down each ratio your campaign needs, then pick the model that covers most of them in one family so that arms differ only in their prompt.
- Only 9:16 and 16:9: Omni or auto is enough.
- Add 1:1: use Kling 3, Seedance, Wan or MiniMax.
- Add 4:5: generate 3:4 and conform with Timeline, or crop with video-filter.
- Mixed set: keep one model for all arms and conform each output to its placement.
Caveat
Model catalogs change. Treat the table as a snapshot of what the repository showed on 2026-10-07, and re-read the live catalog before a large run. We did not render sample clips for this comparison.
Checking a body before you submit
A 400 on aspect_ratio costs nothing but time, yet in a batch it can stall a whole sheet. Read the catalog entry for your model, confirm that the ratio is in supported_aspect_ratios, and confirm that the duration and resolution you want are listed. The /v1/videos catalog controls validation: if the model does not advertise a value, the API returns 400 and does not silently drop it.
This also protects the test. If a ratio were silently ignored, an arm meant to be square could come back wide, and you would only notice when you opened the file.
Most campaigns need two or three ratios from the same concept. Generate the concept once at the ratio that matters most, then decide per placement whether to regenerate natively or conform. Native generation costs another job each time. Conforming costs $0.10 per Timeline minute, so a 6 second conform is $0.10 and has a fixed, known price, but it cannot invent content that was never in the frame.
Sources
Related posts
More in Models
- MiniMax H3 Max on Sume: video, lip-sync and recast, which id to call
Three Sume ids carry the MiniMax H3 Max name: minimax-h3-max for video, a lip-sync route and h3-max-recast. What each takes, its window and its rate.
- Video edit prompt: say what stays, then what changes (Omni Flash)
A prompt pattern for Gemini Omni Flash 1.1 video edit on Sume: one change per pass, an explicit keep clause, and what the video_url route fixes for you.
- Which Sume image models are text-to-image only, with no photo edits
Five Sume image models reject reference images: Higgsfield Soul, Imagen 4 Fast, Imagen 4 Ultra, Recraft V4 and Qwen Image Max. Prices, ratios and edit options.
- An OpenRouter-compatible video API: sume/auto or a pinned model
Sume's POST /v1/videos follows OpenRouter's video generation API field for field. Let sume/auto pick the model, or pin a catalog id like seedance-2.5.
Written by Sume