Create and edit 4K visuals with multiple references, clearer multilingual text, precise instruction following, and flexible formats in one fast image workflow.
One fast model for consistent campaigns, legible copy, and production-ready formats.
Consistent subjects
Carry recognizable people, products, outfits, and visual cues across several references and scenes.
Multilingual copy
Build posters and campaign assets with clearer short-form text in multiple languages.
Precise direction
Translate detailed scene, object, and composition requirements into one organized image.
Flexible delivery
Generate at 1K, 2K, or 4K and compose for a broad range of landscape, square, and portrait ratios.
Key Features
Built for visual systems, not isolated images.
Nano Banana 2 combines Gemini 3.1 Flash Image capabilities with a practical Nanobana workflow for generating, editing, and downloading campaign-ready visuals.
Feature 01 · Multi-reference consistency
Keep the visual identity recognizable from frame to frame.
Use reference images to anchor a person, product, outfit, or art direction while moving the subject into new compositions. That makes connected campaign sets easier to build without recreating the identity each time.
Use up to 14 reference images in the Nanobana generator
Preserve recognizable subjects across new scenes
Combine identity, product, palette, and layout references
Feature 02 · Multilingual text
Design campaign copy directly inside the image.
Stronger text rendering and translation help create posters, signs, menus, and social graphics where short-form copy belongs inside the composition instead of being added as an afterthought.
Render concise headlines and event details
Create localized versions across multiple languages
Keep typography aligned with the surrounding art direction
Feature 03 · Instruction following
Direct a complex scene without losing its structure.
Describe the subject, object count, spatial relationships, camera, lighting, and exclusions in natural language. Nano Banana 2 is designed to follow detailed directions while keeping the result visually coherent.
Place several required elements in one composition
Control relationships such as left, center, behind, or facing
Refine one constraint at a time with focused edits
Feature 04 · 4K and flexible formats
Compose once, then deliver for the placement.
Choose 1K, 2K, or 4K output and work across square, portrait, landscape, panoramic, and extreme aspect ratios. The model can recompose the idea instead of simply cropping the same frame.
Select 1K, 2K, or 4K before generating
Use ratios from tall portrait to ultra-wide panorama
Download the finished result as PNG or JPG
Examples + prompt recipes
See one campaign move across story, type, editorial, and outdoor.
Every example uses the fictional NOVA City Ride identity. Reuse the prompt structure with your own references to test consistency, exact copy, structured layouts, and extreme formats.
A consistent three-scene story
The same rider, outfit, bag, and folding bicycle continue from preparation to riding to transit.
View the prompt+
Using the uploaded rider and bicycle references, create a three-panel editorial commuter story. Panel one shows the coral folding bicycle being unfolded on a quiet sidewalk. Panel two shows the same woman riding through a green city lane. Panel three shows her carrying the folded bicycle on a clean train platform. Preserve her face, black bob haircut, cobalt jacket, gray trousers, coral bag, ivory helmet, and the exact bicycle frame in every panel. Natural morning light, realistic campaign photography, no text.
Exact English and Chinese phrases become part of a restrained, highly usable campaign layout.
View the prompt+
Using the uploaded NOVA rider and bicycle references, design a premium 3:2 bilingual cycling event poster. Preserve the same woman, outfit, coral folding bicycle, and campaign palette. Render only the exact text "NOVA CITY RIDE", "轻装出发", and "12 OCT · 08:00". Use a warm cream background, cobalt type, a coral route line, a clean editorial grid, realistic subject photography, and generous whitespace. No other words or logos.
One direction organizes a hero photograph, product details, a small route graphic, and concise copy.
View the prompt+
Create an open premium magazine spread for the NOVA folding-bike campaign using the uploaded references. On the left, place a large photograph of the same woman riding through a modern city garden. On the right, create four aligned modules for the folding hinge, cream saddle, compact folded silhouette, and a simple coral route diagram. Render only the exact text "NOVA FIELD GUIDE" and "RIDE LIGHT". Warm cream paper, cobalt and coral accents, restrained editorial typography, realistic print texture.
The campaign is recomposed for a very wide station placement while the rider and bicycle stay recognizable.
View the prompt+
Using the uploaded NOVA campaign references, create an ultra-wide metro-station banner. Keep the same woman, face, black bob, cobalt jacket, coral bag, ivory helmet, and coral folding bicycle. Show her riding through a modern city garden from left to right with a coral route ribbon sweeping across the panorama. Render only the exact text "MOVE THROUGH THE CITY" and "NOVA · 08:00". Natural late-afternoon light, premium outdoor campaign photography, strong panoramic balance, no other words or logos.
From a clear brief to a downloadable image in three steps.
Start from text alone, or add references when identity, product shape, composition, or style needs to remain recognizable.
01
Choose your inputs
Start with Text to Image, or upload reference images for an Image to Image task.
02
Set the direction
Describe the subject, layout, lighting, exact text, aspect ratio, and details that must stay unchanged.
03
Generate and refine
Review the result, make one focused change at a time, then download the strongest version.
Nano Banana 2 questions
What is Nano Banana 2?+
Nano Banana 2 is the product name commonly used for Gemini 3.1 Flash Image, an image generation and editing model developed by Google. It combines fast generation with stronger world knowledge, instruction following, text rendering, and reference-image consistency.
What is the relationship between Nanobana, Nano Banana 2, and Google?+
Nano Banana 2 is developed by Google. Nanobana is an independent creative platform that provides a web workflow for supported third-party models. Nanobana does not develop or own Nano Banana 2 and is not affiliated with or endorsed by Google.
Do I need a Google or Gemini account to use it on Nanobana?+
No separate Google or Gemini login is required for this workflow. Sign in to Nanobana, select Nano Banana 2, and use the credits shown in your Nanobana account.
Can I use both text-to-image and image-to-image?+
Yes. Use Text to Image for a new composition from a written brief. Add one or more references for Image to Image when the result should preserve a person, product, palette, layout, or other visual anchor.
How many reference images can I upload?+
The current Nanobana Nano Banana 2 generator accepts up to 14 reference images. Use only references that contribute a clear role, such as identity, product shape, wardrobe, palette, or composition.
Which resolutions, aspect ratios, and file formats are available?+
You can choose 1K, 2K, or 4K output, select from the aspect ratios shown in the generator—including square, portrait, landscape, and panoramic options—and download the result as PNG or JPG.
How do I improve consistency across several images?+
Reuse the same sharp references, name the details that must remain unchanged, and ask for one new scene or edit at a time. Check faces, product geometry, wardrobe, small text, and repeated brand details before publishing.
Does the current Nanobana workflow use Google Search grounding?+
No. The current Nano Banana 2 generation workflow on Nanobana does not enable Google Search grounding. Include any required factual context directly in the prompt or references, and verify factual content before use.
What should I review before publishing a generated image?+
Proofread every word and number, then inspect faces, hands, object geometry, reflections, small labels, and brand details. Generated content can still contain visual or factual mistakes and should receive human review.
Build a complete campaign with Nano Banana 2.
Start from one of the NOVA prompt recipes or upload your own visual references. Nano Banana 2 is already selected in the generator.
