Shortest AI video clip by model: 2, 3, 4 or 5 second floors on Sume
Wan 3.0 starts at 2 seconds; Gemini Omni Flash at 3; Seedance and Kling at 4; MiniMax H3 at 5. The minimum duration for each Sume video model.

The shortest clip you can request on Sume is 2 seconds, and only Wan 3.0 (wan-3.0) goes that low. Gemini Omni Flash 1.1 starts at 3 seconds, the Seedance and Kling 3 rows at 4, and MiniMax H3 and H3 Max at 5. Ask for less than a model's floor and the request is rejected, so the floor decides which model you can use for a quick cutaway, a reaction beat or a sticker-length loop.
Minimum and maximum duration per model
Durations are whole seconds. The table is the capability envelope the catalog publishes; read GET /v1/videos/models before you hard-code any of it, because limits are per model and change when a model row changes.
| Model id | Min (s) | Max (s) |
|---|---|---|
| wan-3.0 | 2 | 30 |
| gemini-omni-flash-1.1 | 3 | 10 |
| seedance-2.5 | 4 | 30 |
| seedance-2, seedance-2-fast, seedance-2-mini | 4 | 15 |
| kling-3 | 4 | 15 |
| grok-imagine-video-1.5 | 4 | 15 |
| minimax-h3, minimax-h3-max | 5 | 15 |
When the floor matters
A 3-second B-roll insert on a 4-second-floor model costs you a trim afterwards. Generating 4 seconds and cutting to 3 works, but you pay for the 4. If the shot is short by design, pin Wan 3.0 or Gemini Omni Flash and request exactly what you need.
Going the other way, a model with a high floor is wasteful for loops and stingers. MiniMax H3 at 5 seconds is a fine length for a product reveal and a poor one for a 2-second logo sting.
Rows that do not follow the pattern
Two catalog rows work from a source video, so the length is not yours to pick. H3 Max Recast takes a source of 5 to 30 seconds and the duration must be the source length, rounded up. Genjutsu Motion Transfer accepts 4 to 30 seconds on the same rule. For both, inspect the input first, then submit its length.
How to request a short clip
Send the integer in duration. A 2-second Wan 3.0 request on the Video Router looks like the sample below; the same field works on /v1/videos.
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: short-001" \
-d '{"model":"wan-3.0","prompt":"A coffee cup steams on a desk, static camera","resolution":"480p","duration":2,"aspect_ratio":"16:9","mode":"async"}'Sources
Related posts
More in Models
- Silent AI video: generate_audio false or drop the audio after
Seedance, Kling and Wan take generate_audio false; Omni and MiniMax H3 always make sound. How to get a silent clip on Sume, and what it costs.
- Sume auto or a pinned video model: what changes past 10 seconds
Auto video defaults to 720p and 8 seconds with 3 to 10 second clips. For 12, 15 or 30 seconds, 1080p or 21:9, pin a model id. How to decide on Sume.
- 9:16 vertical AI video: which Sume models take an aspect ratio
Seedance, Wan 3.0, Kling 3, MiniMax H3 and Gemini Omni Flash take 9:16; Grok Imagine, Genjutsu and H3 Max Recast take no aspect ratio. A per-model table.
- Which AI video models take 1080p on Sume, and which do not
Seedance, Kling, Wan and Omni accept 1080p on Sume; H3 Max refines to it from native 768p; H3, Grok and Genjutsu stop lower. Full matrix.
Written by Sume