Video editing board: Wan 3.0 1,187, Seedance 2.5 1,151; Sume uses Omni
Wan 3.0 tops the AA video-editing board at 1,187 Elo. Sume's documented edit route is video_url on gemini-omni-flash-1.1, third at 1,139. Price inside.

Wan 3.0 is first on Artificial Analysis's video-editing board (AA-Video-Editing v1.1) at 1,187 Elo, and Seedance 2.5 is second at 1,151. Sume's documented edit route is different: you send video_url with gemini-omni-flash-1.1, which is third on the board at 1,139 and $9.64 per minute. Sume's docs describe no edit mode for wan-3.0 or seedance-2.5.
So if your question is how to edit a clip with a prompt through Sume today, the answer is Omni. If your question is which model edits best on a blind vote, the board says Wan 3.0, and you should know Sume does not document that route.
The board
The board ranks editing of a video from a text instruction, with the audio kept. It was last updated in August 2026 when Wan 3.0 arrived.
| Rank | Model | Elo | Listed price per minute | Documented on Sume as an edit route |
|---|---|---|---|---|
| 1 | Wan 3.0 | 1187 | $12.00 | No |
| 2 | Dreamina Seedance 2.5 | 1151 | $40.82 | No |
| 3 | Gemini Omni Flash 1.1 | 1139 | $9.64 | Yes, video_url |
| 4 | MiniMax H3 (768p) | 1119 | $9.60 | No |
| 5 | HappyHorse-1.0 | 1088 | $27.04 | Not listed |
| 6 | Dreamina Seedance 2.0 | 1076 | $27.42 | No |
How the Sume edit call works
The edit goes to the Video Router. video_url is the source clip, the prompt is the instruction, and resolution is optional (the default is 720p). You cannot send aspect_ratio or duration, because the output keeps the source framing. You also cannot combine video_url with image_url, end_image_url or reference lists.
curl -X POST https://api.sume.com/v1/video-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: edit-001" \
-d '{
"model": "gemini-omni-flash-1.1",
"prompt": "Replace the bottle with an apple. Keep everything else the same.",
"video_url": "https://example.com/clip.mp4",
"resolution": "720p",
"mode": "async"
}'What an edit costs
Omni bills the provider list times 1.25 for each output second: $0.125 at 720p, $0.1875 at 1080p. An 8 second source at 720p is 8 x $0.125 = $1.00. Because the edit has no duration field, Sume uses your duration only as a reserve hint, so keep the source short when you test.
Send an Idempotency-Key header with each edit request, as the docs example does, so a retry does not create a second job. Keep the instruction short and concrete, name what must stay unchanged, and compare the result to the source before you publish.
Edits are billed per second of source, so trim the source to the part you want changed before you send it. A 30 second clip with a 6 second change costs $3.75 at 720p, while the trimmed 6 second clip costs $0.75, a saving of $3.00 per edit. Over fifty edits that is $150.00.
Check the first seconds of the output against the source before you accept it. If the instruction asked for one object to change, look at the edges of that object and at anything that moves behind it, and run a second edit only on the frames that failed.
What to do with a #1 you cannot call
A leaderboard rank is a reason to test, not a reason to wait. Run the same edit instruction on Omni at 720p, judge the result against your own acceptance rule, and check which video models Sume lists again when you plan a new project, because the catalog changes.
Sources
Related posts
More in Models
- AI video with native audio: which Sume models let you turn it off
On Sume, only Kling v3 Pro prices audio separately (+$0.07/s). Gemini Omni and MiniMax always make audio; Grok Imagine and Genjutsu make none. Table inside.
- Vidu Q4 takes 15 image refs: what Sume's reference limits are
Vidu Q4 Preview lists up to 15 image and 3 audio references. Sume does not list Vidu; Gemini Omni Flash 1.1 takes 10 images and 3 clips of 3 s, none audio.
- Wan 2.2 CUDA out of memory: offload_model, dtype and t5_cpu flags
The Wan 2.2 README fixes OOM with three flags: offload_model, convert_model_dtype and t5_cpu. Which commands need 80 GB and 24 GB, and what it does not promise.
- Wan 2.2 prompt extension: a DashScope key or a local Qwen model
Wan 2.2's README offers prompt extension via Alibaba DashScope (DASH_API_KEY) or a local Qwen model. What each needs, and what Sume's request takes.
Written by Sume