Nano Banana 2.1 still, then Omni 1080p for 6 s: $1.225 together

A Nano Banana 2.1 still at 1K costs $0.10 on Sume, and animating it with Gemini Omni Flash 1.1 at 1080p for 6 seconds costs 6 x $0.1875 = $1.125, so the pair costs $1.225. The video is 92% of that total.

Google's changelog lists Nano Banana 2.1 as generally available on Oct 6 2026 (read 2026-10-09). On Sume you call it by the id nano-banana-2.1; prices are from the Sume price tool.

Where the money goes

Reroll the still as often as you like: at $0.10 a try, 10 stills cost $1.00, less than one 1080p animation. Settle the still first, then animate.

The two calls

POST /v1/images blocks for up to 30 seconds and returns the image as a Sume-hosted URL in data[0].url. If it answers 202, poll the job before you read the URL. Pass the URL to the Video Router as image_url.

import os, requests
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
img = requests.post("https://api.sume.com/v1/images", headers=H, timeout=60,
    json={"model": "nano-banana-2.1",
          "prompt": "Matte black bottle on marble, soft window light, vertical"})
img.raise_for_status()
if img.status_code != 200:
    raise SystemExit("image job still running: poll its status_url first")
still = img.json()["data"][0]["url"]
vid = requests.post("https://api.sume.com/v1/video-router/generate", headers=H, timeout=60,
    json={"model": "gemini-omni-flash-1.1", "image_url": still,
          "prompt": "Slow orbit around the bottle, light shifts across the label",
          "resolution": "1080p", "duration": 6, "aspect_ratio": "9:16", "mode": "async"})
vid.raise_for_status()
print(vid.json())

Ten assets

A set of ten stills each animated once multiplies the pair price. The table compares two animation resolutions.

Limits and gotchas

Check the image docs for the resolution field before you ask for 2K or 4K stills, since their price is higher than the 1K figure used here.

• Omni takes 3-10 seconds, 16:9 or 9:16. Make the still in the same shape as the clip.
• Audio is always on for Omni, so the animation has a soundtrack.
• A start-frame request uses image_url. For a reference request, use reference_image_urls and a token instead.
• Give the video call its own Idempotency-Key; the image call is not a video job.