A 30-second vertical ad in one request: Seedance 2.5 or Wan 3.0
Seedance 2.5 (4-30 s) and Wan 3.0 (2-30 s) list 9:16 and 1080p on Sume. sume/auto stops at 10 s, so pin a model for a 30 s clip.

A 30-second vertical ad needs no stitching if you pin the model. Seedance 2.5 accepts 4 to 30 seconds and Wan 3.0 accepts 2 to 30, both with 9:16 and 1080p, per Sume's Video Router doc. sume/auto would refuse: it stops at 10 seconds. Thirty seconds is well inside YouTube's 3-minute Shorts limit and TikTok's 10-minute ad limit.
The request
Pin the model, give the ratio, tier and length, and send an idempotency key. If the catalog does not list a value, you get a 400 and no clip.
curl -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: vertical-30s-001" \
-d '{"model":"seedance-2.5","prompt":"A slow push-in on a ceramic mug, steam rising, morning window light","aspect_ratio":"9:16","resolution":"1080p","duration":30}'Seedance 2.5 or Wan 3.0
| Model | Duration | Resolutions | Notes from the docs |
|---|---|---|---|
| seedance-2.5 | 4-30 s | 480p, 720p, 1080p | No bitrate_mode |
| wan-3.0 | 2-30 s | 480p, 720p, 1080p | Text-to-video, image-to-video with end frame, reference-to-video; resolution-tiered list price |
| sume/auto | 3-10 s | 360p-4K | Refuses 30 s |
Drafting and cost control
A 30-second 1080p request is the most expensive way to find out your prompt is wrong. Run the same prompt at 480p and a shorter length first, say 4 seconds on Seedance 2.5, which lists a 4-second minimum, to check the subject and motion. Then raise the duration and the tier.
Check the price before you submit. The Video Router doc lists per-second list rates for some rows, such as Wan 3.0's three tiers, and Sume bills the provider list price times 1.25 at submit, reserved up front. Read the current rate for the row from the catalog, multiply by the seconds you want, and decide whether one 30-second take or two 15-second takes is the better spend.
Remember that a model can also return something unusable. Plan for at least one regeneration in your budget rather than assuming the first take ships.
Tradeoffs
A single 30-second generation holds one scene and one look, which is the point. But if the last second is wrong you pay for all 30 again. A cheap 480p draft at the same length tells you whether the motion works before you spend on 1080p. Neither platform page asks for 30 seconds, so make it as long as the story needs.
Then check the file: width and height against TikTok's 540x960 minimum, and size and bit rate against its 500 MB and 516 kbps limits. The 15-second ad comparison has the shorter version of this choice.
Sources
Related posts
More in Use cases
- A 5-second image-to-video at 1080p: Veo 3.1 $0.40 to $2, Omni $0.94
Five seconds of 1080p costs $0.40 on Veo 3.1 Lite, $0.60 on Fast, $2.00 on Standard and $0.9375 on Omni Flash on Sume, Read 2026-10-06.
- 156 thumbnail options a year on GPT Image 2.5, by quality
52 videos with three thumbnail options each is 156 images. At 1280x720 on Sume that is about $0.62 at low and $5.54 at high.
- 90-second brand film at 1080p: Veo 3.1 from $7.20, Omni on Sume $16.88
Ninety seconds at 1080p is $7.20 on Veo 3.1 Lite, $10.80 on Fast, $36 on Standard and $16.875 on Omni Flash on Sume, before retakes. Read 2026-10-06.
- Abandoned cart reminder: one Sume avatar clip per product, $3.87
Render one 15 second avatar reminder per product with product_image, not per shopper. $3.87 a product at plus, $2.91 at standard. The job body and math.
Written by Sume