PRODUCT UPDATE / Image generation · Game art

Nano Banana 2.1 launches with updates to image editing and consistency

Google released Nano Banana 2.1 on October 6. The image model supports up to 14 references and 1K, 2K and 4K output; Scenario added it on October 7.

Published Event
Three Nano Banana 2.1 design samples: a Zurich Modern website, a bucket-hat poster and a blue Tundra website mockup.
Website and poster layouts.
Who should pay attention

A candidate for iterating on game art references, interface mockups and promotional visuals.

Google released Nano Banana 2.1 in its October 6 Gemini API update. Google says the model improves visual design, localized editing, subject consistency and text rendering. Those capabilities are relevant to teams revising game art references, interface mockups and promotional images.

Mark a subject and change its surroundings

The new Google DeepMind showcase includes an input image with a hand-drawn outline around a water-covered dandelion seed. The example prompt asks for different environments while keeping the subject's position and size.

An outlined dandelion seed in the input image and four results placing it on a mossy log, an autumn leaf, a beach and a tiled table.
An outlined subject in different surroundings.

A second example places the same peach shoulder bag in a geometric studio, a terrazzo setting and a cloud scene. These are Google's editing and consistency demonstrations. When making character or prop variations, check the silhouette, material details and lighting against the project's requirements.

An input peach shoulder bag and three variations in a geometric studio, a terrazzo setting and a cloud scene.
One shoulder bag in three settings.

Up to 14 references and wide aspect ratios

The model documentation lists up to 14 reference images and 1K, 2K and 4K output, with 1K as the default. Google also reports fixes for tiling artifacts at 2K and 4K in the 1:4, 4:1, 1:8 and 8:1 formats.

Wide formats offer a way to draft scene backdrops, while multiple references can supply characters, objects and style information together. Backgrounds intended to tile still need a separate seam check.

Google API costs and Scenario access

The Gemini API model ID is gemini-nano-banana-2.1. Standard API pricing charges for image-output tokens. Google's square-image examples work out to about $0.0336 for 1K (1024 × 1024), $0.0504 for 2K (2048 × 2048) and $0.113 for 4K (4096 × 4096). These amounts cover image output only; inputs, text and thinking output, and searches are charged separately. The model has no free API tier.

Scenario added Nano Banana 2.1 on October 7. Its official RSS announcement lists generation and editing, up to 14 reference images and panoramic formats up to 8:1. The dollar figures above are Google's API prices and do not describe Scenario credit charges.

Sources (6)
  1. Google Gemini API · October 6, 2026 release notes
  2. Google Gemini API · Nano Banana 2.1 model specifications
  3. Google DeepMind · Introducing Nano Banana 2.1 and original examples
  4. Google Gemini API · Nano Banana 2.1 Standard and Batch pricing
  5. Scenario · October 7, 2026: Nano Banana 2.1
  6. Scenario · Official changelog RSS and October 7 announcement details

No sponsorship or affiliate links in this article.