Vidu Q4 Preview makes 16 s clips: which Sume rows go past 10 s?

Vidu Q4 Preview generates clips of 3 to 16 seconds, per its image-to-video and reference-to-video API pages. A 16-second clip is possible on Sume with two catalog rows: Wan 3.0, which accepts 2 to 30 seconds, and Seedance 2.5, which accepts 4 to 30. Kling 3, MiniMax H3 and the Seedance 2.0 family stop at 15 seconds, and Gemini Omni Flash 1.1 stops at 10.

Duration ranges

Sume's durations are whole seconds, and the allowed range is per model. Read supported_durations from GET /v1/videos/models rather than relying on one number. A duration outside the range is a 400 unsupported_capability, not a silent clamp.

What a 16 second clip costs on Sume

Wan 3.0 is priced per second by resolution: $0.0625 at 480p, $0.125 at 720p and $0.25 at 1080p. A 16-second clip is therefore $1.00 at 480p, $2.00 at 720p and $4.00 at 1080p. Seedance 2.5 is priced per video token, so its 16-second price depends on resolution and shape, and you should read it from the catalog row instead of estimating.

Request a 16 second clip

Steps:

• Set model to wan-3.0.
• Set duration to 16 and pick a resolution.
• Add a first_frame if you want to pin the opening shot.
• Poll the job and download from unsigned_urls[0].

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"wan-3.0","prompt":"A courier cycles through a night market, handheld","duration":16,"resolution":"720p","aspect_ratio":"9:16"}'

Longer is not always better

Length changes cost and risk. A 16-second Wan clip at 1080p is $4.00 on Sume, and if the last four seconds drift off-model you pay for them anyway. Two 8-second jobs at the same resolution cost the same, $4.00, but you can reject one and re-run it for $2.00.

So choose a single long job only when the shot is continuous, such as a camera move through one space. For anything with a cut, generate separate clips and join them on the timeline. The shorter units also let you mix rows, using Omni for a 4K insert and Wan for the wide shot.

Planning around the longest clip

If a shot needs more seconds than the row allows, cut it in editing rather than asking for a longer clip. Plan shots at the length you need, add a second of handle at each end for trims, and check that the row's minimum fits the shortest shot. Seedance 2.x and Kling 3 start at 4 seconds, H3 at 5 and Omni at 3, so a 2 second insert is only available on Wan 3.0, whose range starts at 2.

What Sume does not do

Sume does not list a Vidu model, and Wan 3.0 is a different model with its own look, so a 16-second Wan clip is not a Vidu clip. Wan 3.0 also tops out at 1080p, while Vidu lists up to 4K. Pick by the shot, not only by length.

Also note that the Vidu press coverage and its API page are not identical on the reference-to-video minimum, so use the API page's 3 to 16 seconds when you plan.