Black Friday 2026 banner pack in six ratios from one Nano Banana 2 run

Black Friday 2026 is Friday, November 27, and one Sume model covers a whole banner pack without a crop step: Nano Banana 2 accepts 1:1, 4:5, 9:16, 16:9, 21:9 and 4:1 in the same catalog row. Six banners, one per ratio, cost $0.60 at the 1K rate of $0.10, and three retakes each bring that to $1.80.

The ratio list is read from GET /v1/images/models (read 2026-10-03). The 4:1 strip is the constraint: among all 19 catalog rows only Nano Banana 2 lists 4:1, 1:4, 8:1 and 1:8.

Which models take each ratio

Counts are how many of the 19 catalog models list the ratio in their aspect_ratio descriptor. A request with a ratio outside the list is rejected with 400 unsupported_parameter, so a one-model pack avoids a fallback branch.

Run the six calls

Keep the prompt identical across ratios except for layout words, so the six banners read as one campaign. Pass the same product photo as the reference in every call and set aspect_ratio explicitly; resolution accepts 512, 1K, 2K and 4K, and higher tiers cost more than the 1K figure used here.

Put the offer text in code, not in the model: the safest banner leaves a clear area where your own template adds the price and the date, because wrong digits in a sale banner are expensive.

{
  "model": "google/nano-banana-2",
  "prompt": "Dark studio product shot, space on the left for text, no words in the image",
  "aspect_ratio": "4:1",
  "resolution": "1K",
  "input_references": [
    {"type": "image_url", "image_url": {"url": "https://example.com/product.png"}}
  ]
}

When to pick another model

If every banner needs rendered text, Ideogram 4.5 is the text-focused row but lacks 21:9 and 4:1; if the pack needs 16 references, ChatGPT Image 2.5 does not list 21:9. Pick the model for the hardest banner, then check the others against its ratio list.