Which Sume video row takes 1:1, 3:4 or 21:9? Aspect ratio matrix
Seedance and MiniMax rows take 21:9; Wan 3.0 skips it; Kling 3 stops at 16:9, 9:16 and 1:1; Omni and Veo are 16:9 or 9:16 only. The full matrix by row.

Seedance 2.5, Seedance 2 and the MiniMax H3 rows take 1:1, 3:4 and 21:9. Wan 3.0 takes 1:1 and 3:4 but not 21:9. Kling 3 stops at 16:9, 9:16 and 1:1, and Omni and the Veo rows allow only 16:9 and 9:16.
The matrix below comes from the capability table in Sume's Video Router catalog code, and the Video Router docs describe the same model vocabulary.
The matrix
A tick means the ratio appears in the row's capability entry. auto and adaptive also appear for the Seedance, Wan and MiniMax rows and let the provider choose, so they are not listed as columns. The Grok row lists no aspect ratios, because it follows the input image.
Veo rows come from Sume's Veo catalog entry, and they appear only where your workspace lists them.
| Row | 16:9 | 9:16 | 1:1 | 4:3 | 3:4 | 21:9 |
|---|---|---|---|---|---|---|
| seedance-2.5 and seedance-2 | Yes | Yes | Yes | Yes | Yes | Yes |
| minimax-h3 and minimax-h3-max | Yes | Yes | Yes | Yes | Yes | Yes |
| wan-3.0 | Yes | Yes | Yes | Yes | Yes | No |
| kling-3 | Yes | Yes | Yes | No | No | No |
| gemini-omni-flash-1.1 | Yes | Yes | No | No | No | No |
| veo-3.1 rows | Yes | Yes | No | No | No | No |
| grok-imagine-video-1.5 | Input image | Input image | Input image | Input image | Input image | Input image |
Ratios nobody on Sume lists
None of these capability entries lists 4:5, 5:4 or 9:21. If you need a 4:5 feed ad, generate at 3:4 or 1:1 and crop, as the linked 4:5 post explains. The docs' general ratio list also names 3:2, 2:3 and 9:21, but those are not in the rows above, so read supported_aspect_ratios for the row you pick instead of the general list.
A request with a ratio the row does not list fails validation, which costs you a failed call, not a charge.
Picking by shape
For ultra-wide, use Seedance 2.5 or a MiniMax H3 row. For a square social tile, any row with 1:1 works, and Wan 3.0 at 480p is among the cheapest at $0.0625 per second, the same as MiniMax H3 at 480p. For portrait feed at 3:4, Seedance, MiniMax and Wan all work. For Omni's 4K, you are limited to 16:9 or 9:16.
Kling 3 deserves a note: its catalog entry lists 1080p as the only resolution, with 16:9, 9:16 and 1:1. A 720p request from older code is accepted but renders at 1080p.
- 21:9 trailer shot: Seedance 2.5 or MiniMax H3.
- Square tile: Wan 3.0 or Kling 3.
- Vertical 4K: Gemini Omni Flash 1.1.
- Match the input image: Grok row.
Always read the live list
Capability tables change. GET /v1/videos/models returns supported_aspect_ratios per row, and that list is the contract. The /v1/videos catalog filters to the ratios the OpenRouter-compatible contract allows, so a Video Router-only value such as adaptive may not appear there. Build your request from the live list at run time.
Cost of choosing the wrong shape
Choosing a ratio late is the expensive path. If you generate a 9:16 clip and later need it at 1:1, you either crop away more than half the frame or you regenerate. Decide the delivery shape first, then pick the row. For a campaign that needs both 16:9 and 9:16 from one idea, send two requests from the same prompt rather than cropping one output.
Per second, a 5-second 720p Wan 3.0 clip bills $0.625 at any ratio the row lists, so the second request doubles the cost to $1.25. A Seedance 2.5 pair at 720p bills $5.778 for two 5-second clips. Plan both shapes in the first draft.
Sources
Related posts
More in Models
- Chinese AI video models by API: Kling, Wan, MiniMax limits and prices
Which Chinese video models can you call on Sume? kling-3, wan-3.0, minimax-h3 and minimax-h3-max: clip length, resolution, audio, references and price a second.
- Clef-omni reads video: a cheap check before a second video job
Cloudflare's Clef-omni takes audio and video input at $0.15 per million tokens. Sume does not offer Clef, so this is a pre-check pattern to run outside Sume.
- Eleven v4 on Sume lists 44 language codes, not 90+
ElevenLabs says v4 speaks 90+ languages. Sume's TTS Router catalog lists 44 ISO codes for eleven-v4, and 28 for Multilingual v2. Check yours before you send.
- FastH3 Trim: MiniMax H3 on an 8 GB GPU, and who may use it
FastH3 Trim prunes MiniMax H3 to 42 blocks and 8 steps. File sizes, the 8 GB claim, the quality trade, and what the H3 Community License allows.
Written by Sume