Back to Models Guide
Googleimage

Complete Guide to Using Nano Banana 2

Powered by Gemini 3.1 Flash with multi-image input, Google Search grounding, and up to 4K resolution.

Try This ModelTutorial

Overview

Nano Banana 2 is built on Google's Gemini 3.1 Flash architecture, bringing multimodal understanding to image generation. It accepts up to 14 reference images alongside a text prompt, enabling powerful composition, style matching, and character consistency across generated outputs.

A standout feature is Google Search grounding, which allows the model to incorporate real-time information from the web into its generations. This is especially useful for creating images of current events, trending styles, or specific real-world locations and products.

The model supports output resolutions up to 4K and offers a wide selection of aspect ratios including ultrawide 21:9. Combined with strong text rendering inherited from the Gemini architecture and flexible output format options, Nano Banana 2 is one of the most versatile image generation models available on the platform.

Capabilities

  • Multi-image input support with up to 14 reference images
  • Google Search grounding for real-time information integration
  • Output resolution up to 4K for print-quality results
  • Strong bilingual text rendering in generated images
  • Wide aspect ratio selection including ultrawide 21:9
  • Character and style consistency across generations using reference images

Use Cases

1

Creating consistent character illustrations across a series

2

Generating images informed by current events or real-world references

3

High-resolution artwork for print and large-format display

4

Brand asset generation with style-matched reference images

5

Multi-reference compositions that blend elements from several source images

Input Parameters

prompt
textarearequired

A text description of the image you want to generate (max 20000 chars).

Describe the desired image in detail. You can reference the uploaded images contextually, e.g., 'Create a portrait in the style of the first image, featuring the person from the second image'. Maximum 20000 characters.

Input Images
file

Input images to transform or use as reference (JPEG/PNG/WebP, ≤30MB, up to 14).

Upload up to 14 reference images. These can serve as style guides, character references, or compositional templates. The model will intelligently incorporate visual elements from these references.

aspect_ratio
select

Aspect ratio of the generated image.

Choose from 15 aspect ratio options. 'Auto' lets the model decide based on content. Use specific ratios when you need precise framing for a particular platform or layout.

Options
Auto1:11:41:82:33:23:44:14:34:55:48:19:1616:921:9
Default: auto
google_search
toggle

Use Google Web Search grounding to generate images based on real-time information.

Enable this to ground the generation in real-time web data. Useful for generating images of specific products, locations, or current cultural references.

Default: false
resolution
select

Resolution of the generated image.

1K is fast and suitable for previews. 2K balances quality and speed. 4K is best for final output intended for print or large displays.

Options
1K2K4K
Default: 1K
output_format
select

Format of the output image.

JPG is smaller and faster for web use. PNG preserves quality and supports transparency.

Options
JPGPNG
Default: jpg

Tips & Best Practices

Multi-image consistency technique
Google Search grounding for accuracy
Resolution ladder approach
Leverage the long prompt limit

Related Models

Google Imagen 4

GoogleView Guide →

Google Imagen 4 Fast

GoogleView Guide →

Google Nano Banana Edit

GoogleView Guide →