Nano Banana 2.1
Google's Nano Banana 2.1 creates images from text and edits existing images through conversation. It offers 1K, 2K, and 4K output, reference-image guidance, clearer lettering, and more consistent characters across edits.
Modalities
In / Out Price
30% off$1.05/ 1M
$1.5 Official price
$5.25/ 1M
$7.5 Official price
Context
131.072K
Released
Oct 15, 2026
Playground
Output
Explore AI StudioAPI call examples
POST /v1/images/generationscurl --request POST 'https://api.tokatlas.com/v1/images/generations' \
--header "Authorization: Bearer $UNIROUTER_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "gemini-nano-banana-2.1",
"prompt": "A serene koi pond at sunset.",
"size": "1024x1024",
"n": 1
}'Providers & routes
Current provider · Google
Platform configured price · Group: default
USD / 1M tokens. Each row shows input and output from the same tier. Displayed rates use the lowest available group multiplier; your API key group may differ.
Nano Banana 2.1 for reference-based and iterative edits
Use reference images for iterative edits. Describe the target area and the details that should carry through to the next revision.
Multiple references
Identify the sources for products, people, and layout so their constraints do not conflict.
1K to 4K
Review composition at a supported draft size before increasing resolution.
Local revisions
Change one specified area per round while preserving subject identity and crop.
Layout and lettering
Set the information hierarchy for a poster, then check each heading and label.
About
Composing from multiple references
Nano Banana 2.1 supports generation, editing, and repeated revisions. Listed upstream capacities include up to 14 references: up to ten objects and four characters. Give each a role such as character identity, product shape, or lighting.
Size, aspect ratio, and input types
Upstream sizes include 1K, 2K, and 4K. The reference also lists video and PDF input, depending on the endpoint. For unusual aspect ratios, review composition at low resolution before checking the edges at delivery size.
Comparing with Pro and other image models
Use the same product asset, prompt, and aspect ratio to compare product consistency after edits, lettering accuracy, and time to output. Use the results to decide whether drafts and final edits need different models.
For successive edits, give the original reference and most recent approved image distinct roles, with explicit keep and change lists. Arrange text-heavy visuals and product labels so the final wording can be checked individually.
Nano Banana 2.1 FAQ
How should I prompt with multiple reference images?
Give each image a purpose, such as using the first for the product and the second for color. Start with a small reference set and add more only as needed.
How should I prototype a high-resolution asset?
Choose the intended aspect ratio and an available size, then validate framing on a smaller output. Raise resolution for the final pass and inspect text and important details.
Does an image without a visible logo have no watermark?
Google embeds invisible SynthID in generated images. Visually checking for an overlay is separate from checking for embedded provenance.
When should I compare against Nano Banana Pro?
Start with 2.1 for routine edits and iterations. Compare Pro using the same references when the layout has dense type, several languages or strict composition rules.