18-second AI video on Sume: Seedance 2.0 stops at 15, 2.5 is $10.40

For an 18-second clip at 720p, use seedance-2.5 on Sume: 18 x $0.5778 = $10.4004, shown as $10.40. seedance-2 cannot take an 18-second request because its duration range is 4 to 15 seconds. The two rows share the same modes (text, image, first and last frame, reference images, video and audio), so the choice comes down to length and rate.

Side by side at 720p

The 15-second row is the longest clip both models accept, so it is the fair comparison. The 18-second row exists only for Seedance 2.5.

Is a 15 second clip plus a 3 second clip cheaper?

You could split the shot: 15 seconds on Seedance 2.0 ($5.67) plus the 4-second minimum on Seedance 2.0 ($1.51) gives 19 seconds of footage for $7.18, which is cheaper than one Seedance 2.5 job. The cost is a seam. Two independent generations rarely match in lighting, subject position and motion at the cut, and you would have to join them on a timeline.

If the action must be continuous, pay for Seedance 2.5. If the second half can be a different angle, two Seedance 2.0 jobs save money.

• A request with duration above the model's maximum is refused; read supported_durations from GET /v1/videos/models before you submit.
• Last-frame chaining is possible: both rows accept first_frame and last_frame in frame_images.
• Resolution options are the same: 480p, 720p and 1080p.

Checking the limit in code

Do not hard-code 15 or 30. The models endpoint returns the list for each model, and a small helper can choose the cheapest row whose list contains your duration.

import os, requests

headers = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
r = requests.get("https://api.sume.com/v1/videos/models", headers=headers, timeout=30)
r.raise_for_status()
want = 18
for model in r.json()["data"]:
    if model["id"].startswith("seedance") and want in model["supported_durations"]:
        print(model["id"], "accepts", want, "seconds")

A rule of thumb for choosing

Choose by the longest continuous shot you need. Up to 15 seconds, Seedance 2.0 is about 35 percent cheaper per second at 720p ($0.378 against $0.5778, a ratio of 0.654). From 16 to 30 seconds only Seedance 2.5 or Wan 3.0 can take the job in one request.

Wan 3.0 is the third option for 18 seconds: 18 x $0.125 = $2.25 at 720p. It is a different model with a different look, so run a short draft on each before you decide. Price should not be the only filter when a client has approved a style.