Omni 9:16 and 16:9 360p drafts of one prompt: 75 cents for the pair

Drafting one prompt in both of Gemini Omni 1.1 Flash's aspect ratios at 360p costs about $0.75 for two 10-second clips on Sume, and $0.225 for two 3-second clips. Omni on Sume takes 16:9 or 9:16 and nothing else, so one pair covers every aspect ratio the model offers. Look at both before you pay for a 720p, 1080p or 4K final in either.

Why a pair, not a crop

A model that frames a vertical shot composes it differently from a landscape one. A subject that sits well in 16:9 can be cut by a 9:16 crop. Timeline 1.0 can fit a 16:9 source into a vertical frame with cover (crop), contain, stretch or blur (Sume docs: Timeline 1.0, read 2026-10-05), but each option either crops or letterboxes. The aspect ratio belongs in the prompt request, not in post.

Google positions the 360p mode for rapid prototyping and storyboard iteration, at a third of the cost of the standard 720p mode (Google, read 2026-10-05). A two-aspect pair is an obvious use for it.

Cost of the pair

Sume bills the provider list times 1.25 per output second, by resolution. At the list rates in Sume's estimator when this was written ($0.03 per second at 360p and $0.10 at 720p) the Sume rates are $0.0375 and $0.125 per second. Read GET /v1/videos/models and the API pricing page for the live rates before you rely on these.

Request shape

Send the same body twice and change only aspect_ratio. Use distinct Idempotency-Keys, because the bodies differ (a reused key returns a 409). Sume reserves the estimated amount at submit and captures it on success; a failed job is refunded.

import os, requests

AUTH = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}

for ratio in ("9:16", "16:9"):
    body = {
        "model": "gemini-omni-flash-1.1",
        "prompt": "A barista hands over a latte, steam, shallow depth of field",
        "resolution": "360p",
        "duration": 10,
        "aspect_ratio": ratio,
    }
    r = requests.post("https://api.sume.com/v1/videos", json=body,
                      headers={**AUTH, "Idempotency-Key": f"latte-draft-{ratio}-v1"})
    r.raise_for_status()
    print(ratio, r.json()["id"])

Pick the winner, then finish once

Decide on composition and motion, not sharpness: 360p cannot show fine text, small logos or skin detail, so check those separately on a final. If both aspect ratios matter for the campaign, finish both, which doubles the final cost; if one platform leads, finish one. A 16:9 and a 9:16 final of the same prompt are separate generations, so expect the two to differ in details beyond the crop.

For a longer cut, repeat the pair per shot and join the winners on a timeline. The timeline output defaults to 1080x1920, so a 16:9 winner needs fit set deliberately, and a vertical-first plan should draft in 9:16 only.

What a 360p pair can and cannot tell you

Use the pair to judge framing: where the subject sits, whether the camera move reads in a tall frame, whether anything important falls outside the crop. It will not show whether small text is legible, whether a logo survives, or how skin looks. Those need the resolution you will publish at. Sume lists 360p, 720p, 1080p and 4K for Omni, so the ladder is wide enough to test each concern at the cheapest tier that can show it.

Keep both drafts and the final decision in your records. If you later change the campaign from vertical to landscape, you will know the 16:9 draft already worked and can skip a round.