Ultra-wide title cards: five 6 s Seedance 2.5 clips in 21:9 at 1080p

A 6-second 21:9 Seedance 2.5 clip at 1080p costs $8.57 on Sume, only $0.04 more than the same clip at 16:9. Ultra-wide is a real option for title cards and cinematic bars, and the price premium is negligible (read 2026-10-09).

Ultra-wide costs about the same

A film title card, a festival bumper or a trailer slate benefits from the wide cinema look of 21:9. Seedance 2.5 lists 21:9 first among its six ratios. Wan 3.0 does not list 21:9, so the ratio decides the model.

Request

Draft the card at 480p first; the ratio is the same, only resolution changes.

import os, time, requests

API = "https://api.sume.com/v1/videos"
H = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"]}
body = {"model": "seedance-2.5", "prompt": "Dark ultra-wide title card: embers drift across black, a silhouette of a lighthouse resolves in the distance", "duration": 6, "resolution": "1080p", "aspect_ratio": "21:9"}
job = requests.post(API, headers={**H, "Idempotency-Key": "title-card-219-001"}, 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 trailer with five title cards

Five 6-second cards at 21:9 and 1080p cost 5 x $8.57 = $42.85. The 480p draft of all five is $8.06, and the 720p version $17.42.

Drafting at 480p and finishing only the three that survive the cut costs $33.77. Finishing all five at 1080p without drafts costs $42.85.

Remember that no v1 model accepts a seed, so the 1080p finish is a new generation rather than a re-render of the draft. Treat the draft as a concept test, not a preview of the exact frames.

Check before you spend

• The 480p draft of the same card costs $1.61, so four drafts are still cheaper than one 1080p render.
• Read supported_aspect_ratios from GET /v1/videos/models before sending; an unsupported ratio is rejected with the supported list in the details.
• Cropping bars onto a 16:9 render throws away picture height; generating at 21:9 asks for a wide composition from the start.
• Title text is better added in an edit step than asked from the model.