Autumn menu photo refresh: 8 dishes, 3 takes on Nano Banana 2

Eight dish photos with a seasonal garnish edit, three takes each, cost $2.40 on Nano Banana 2 at 1K via Sume. Delivery apps may require unedited photos.

4 min readSume
All posts

Eight dish photos with a seasonal edit (a sprig, a dusting of spice, an autumn napkin) and three takes each cost $2.40 on Nano Banana 2 through Sume, at the 1K rate of $0.10 per image. Nano Banana 2 accepts aspect_ratio: "auto", so each dish keeps its crop.

Before you edit a menu photo, check where it will be used: some delivery platforms ask for unedited photos of the real dish, and an edited photo that shows food you do not serve is a customer-trust problem. The cost numbers are from the Sume catalog (read 2026-10-03).

Cost

Resolution tiers are 512, 1K, 2K and 4K, and higher tiers cost more than the 1K figure, so keep drafts at 1K or 512 and step up only for the final.

Autumn menu refresh on Nano Banana 2 at 1K (read 2026-10-03)
ScenarioImagesTotal
One take for 8 dishes8$0.80
Three takes per dish24$2.40
Three takes per dish at 12 dishes36$3.60

Where edits are safe and where they are not

Safe: backdrop, napkin, light color, garnish that you actually serve in autumn. Unsafe: portion size, ingredients, anything that would change what a customer receives. Write that rule into the prompt ("do not change the food itself") and into your review step.

For delivery-app listings, keep the original photo as the listing image where the platform's rules require it, and use the edited version on your own site and social posts.

  • Use the real dish as the reference every time.
  • Review the output at full size for added or missing ingredients.
  • Label edited photos where the platform or local law asks.

One dish

The reference is the original photo; the prompt names only the change.

{
  "model": "google/nano-banana-2",
  "prompt": "Keep the dish unchanged. Add an autumn linen napkin and two oak leaves beside the plate, warm window light",
  "aspect_ratio": "auto",
  "resolution": "1K",
  "n": 3,
  "input_references": [
    {"type": "image_url", "image_url": {"url": "https://example.com/dish-3.jpg"}}
  ]
}

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume