Vertical recipe short on Wan 3.0: 16 s with a plated-dish last frame

A 16-second 9:16 Wan 3.0 recipe clip costs $1.00 at 480p, $2.00 at 720p and $4.00 at 1080p on Sume. Wan 3.0 pricing is per second by resolution with no aspect-ratio term, so vertical costs the same as landscape (read 2026-10-09).

Price around 16 seconds

A recipe short needs enough time for three beats: ingredients, a pour or chop, and the plated dish. Sixteen seconds fits that, and Wan 3.0 accepts the length as a plain whole number.

Use a plated-dish photo as the last frame

Wan 3.0 supports first and last frames. A last frame of the real dish pulls the clip toward the finished plate; a last frame cannot be sent without a first frame, so send both.

import os, time, requests

API = "https://api.sume.com/v1/videos"
H = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"]}
body = {"model": "wan-3.0", "prompt": "Top-down: tomatoes are chopped, tossed in a pan, spooned over pasta", "duration": 16, "resolution": "720p", "aspect_ratio": "9:16", "frame_images": [{"type": "image_url", "image_url": {"url": "https://example.com/board.jpg"}, "frame_type": "first_frame"}, {"type": "image_url", "image_url": {"url": "https://example.com/plated.jpg"}, "frame_type": "last_frame"}]}
job = requests.post(API, headers={**H, "Idempotency-Key": "recipe-pasta-001"}, json=body, timeout=60).json()
deadline = time.time() + 1800
while time.time() < deadline:
    time.sleep(30)
    job = requests.get(job["polling_url"], headers=H, timeout=60).json()
    if job["status"] == "completed":
        print(job["unsigned_urls"][0], job["usage"]["cost"])
        break
    if job["status"] in ("failed", "cancelled"):
        print(job["status"], job.get("error"))
        break

A week of recipe shorts

Seven recipe clips at 16 seconds and 720p cost 7 x $2.00 = $14.00. At 480p the week is $7.00, and at 1080p $28.00.

If a recipe needs 20 seconds, the clip costs $2.50 at 720p. If 12 seconds is enough it costs $1.50. The step is $0.125 per second, so you can fit the length to the recipe.

For consistency across the week, keep one reference image of the kitchen set in input_references for every clip. Wan 3.0 accepts up to 10 reference images in the catalog.

What to check

• Vertical placements play at phone size; 720p is usually the stopping point before the cost of 1080p ($4.00) doubles.
• If both frame_images and input_references are sent, frame_images controls the mode and the request is image-to-video.
• Add the ingredient list as a caption instead of in the prompt.
• Wan 3.0 accepts video and audio references too, so a voiceover reference can set the pace.