60-second AI video API: two 30 s jobs or four 15 s jobs on Sume

A 60-second AI video on Sume takes at least two jobs, because no model goes beyond 30 seconds. The cheapest route is two Wan 3.0 jobs of 30 seconds: $3.76 at 480p, $7.50 at 720p and $15.00 at 1080p.

The alternative is four 15-second jobs on a 15-second model, which gives you more cut points and, on Kling 3, costs $8.40 with audio off.

What does each way of building 60 seconds cost?

Sume bills the provider list times 1.25 and rounds each job up to the cent, so the totals below are job totals added together.

Is two big jobs better than four small ones?

One seam is easier to hide than three. Two 30-second jobs give a single cut at the midpoint and keep both halves on a model that takes reference images. Four 15-second jobs suit a piece with natural scene changes every 15 seconds, such as a four-step explainer.

Kling 3 takes no reference images, so four Kling jobs rely on the first frame to carry continuity. Wan, Seedance and the MiniMax rows accept input_references for the same subject across jobs.

How do you keep the look consistent?

Pass the same reference images to every job, or chain the jobs by starting each from the final frame of the one before. The request below uses the first approach with Wan 3.0.

• Same model, resolution and aspect ratio on every job.
• Reuse the same input_references entries across jobs.
• Use an Idempotency-Key per job so a retry never bills twice.

What does the request look like?

Two jobs of 30 seconds each, keys read from the environment.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
ref = {"type": "image_url", "image_url": {"url": os.environ["REF_URL"]}}
scenes = ["Opening: she enters the workshop",
          "Closing: she finishes the piece and smiles"]
for i, scene in enumerate(scenes):
    r = requests.post(
        "https://api.sume.com/v1/videos",
        headers={**H, "Idempotency-Key": f"film-60-{i}"},
        json={"model": "wan-3.0", "prompt": scene, "duration": 30,
              "resolution": "720p", "input_references": [ref]},
    )
    r.raise_for_status()
    print(i, r.json()["id"])

What does it cost to test first?

A 5-second 480p draft is $0.32 on Wan. Two drafts cost $0.64, well under 10 percent of the $7.50 final.