40-second AI video API: a 30 s job plus a 10 s job on Sume

You cannot request a 40-second clip from any single Sume video model, because the longest duration is 30 seconds. Make one 30-second job and one 10-second job and join them. On Wan 3.0 that costs $2.51 at 480p, $5.00 at 720p and $10.00 at 1080p.

The 10-second half can also run on Gemini Omni Flash 1.1, Kling 3, Grok Imagine Video 1.5 or MiniMax, since all of them accept 10 seconds.

What does the 30 + 10 split cost?

Sume bills the provider list times 1.25 and rounds each job up to the cent. Wan 3.0 is $0.0625, $0.125 and $0.25 per second billed at 480p, 720p and 1080p.

Why split 30 + 10 rather than 20 + 20?

The price is the same either way, because Wan bills by the second. The split changes the creative plan. A 30-second opening carries the main scene and a 10-second closer gives you a short beat for a product shot or call to action.

A 20 + 20 split gives you two equal scenes. Choose by the script, not the price.

How do you make the join invisible?

Start the second job from the last frame of the first. Pull the final frame from the first video, upload it, and send it as a first_frame in frame_images. Repeat the scene description in the prompt so lighting and framing stay the same. If both frame_images and input_references are sent, frame_images wins and the job runs as image-to-video.

• Keep the same model and resolution for both jobs.
• End the first prompt on a calm, held shot so the last frame is clean.
• Join the two files in your editor; Sume returns one file per job.

What does the request look like?

Two POST /v1/videos calls, each with its own idempotency key.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
parts = (("promo-40-a", 30, "Tour of a ceramics studio, slow tracking shot"),
         ("promo-40-b", 10, "Close-up of a finished mug on a shelf"))
for key, secs, prompt in parts:
    r = requests.post(
        "https://api.sume.com/v1/videos",
        headers={**H, "Idempotency-Key": key},
        json={"model": "wan-3.0", "prompt": prompt,
              "duration": secs, "resolution": "720p"},
    )
    r.raise_for_status()
    print(key, r.json()["id"])

Should you draft first?

Yes. A 5-second 480p test of each half costs $0.32 per job and shows whether the prompt works before you commit $5.00 to the pair.