Skip to content

Nano Banana AI Image Generator

Google DeepMind's Gemini 2.5 Flash Image — the fast, affordable everyday image model for text-to-image and natural-language editing.

Supports:
Text to ImageImage to Image

Features of Nano Banana

Natural-Language Image Editing

Just describe the change. Remove a passerby, swap an outfit, relight the scene, or fix a small detail — Nano Banana preserves the rest of the photo while applying the edit you asked for.

Multi-Image Fusion

Upload up to three reference images to blend subjects, keep a character or product consistent across scenes, or combine a style image with your subject. Each reference adds context the model reasons over — not just a style filter.

Built for Speed and Volume

Nano Banana sits on the Flash tier for a reason: single-digit second generation, low cost, and generous throughput. It's the right default when you need dozens of variations or a quick first draft.

11 Aspect Ratios Out of the Box

Go from square social posts to cinematic 21:9 hero banners with one parameter. Pick a ratio, or leave it on Auto to let the model choose based on your prompt.

Safe by Default with SynthID

Every Nano Banana output carries an invisible SynthID watermark signed by Google DeepMind. The watermark doesn't affect image quality but makes provenance verifiable — important for publishers, marketplaces, and compliance-sensitive teams.

Text-to-Image and Image-to-Image in One Model

The same model handles prompt-only generation and multi-reference editing. Start from nothing, start from a product photo, or start from a sketch — the pipeline and pricing stay the same.

Credits Cost

Flat pricing — no resolution tiers

3

credits per image

How to Use Nano Banana

From prompt to finished image in three steps

01

Write a Clear Prompt

Describe the subject, setting, and mood. For edits, drop in up to 3 reference images and say what to change.

02

Pick an Aspect Ratio

Choose one of the 10 ratios, or leave it on Auto and let the model match your prompt.

03

Generate in Seconds

Most images finish in under 10 seconds. Iterate by refining the prompt — each try is only 3 credits.

Technical Specifications

10
Reference Images
Model Provider
Google DeepMind
Model Name
Gemini 2.5 Flash Image (Nano Banana)
Release
Preview 2025-08-26 · GA 2025-10-02
Aspect Ratios
1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 5:4, 4:5, 21:9, Auto
Output Format
PNG
Provenance
SynthID watermark embedded in every output
Typical Speed
Under 10 seconds per image
Supported Scenes
Text-to-Image, Image-to-Image

*The specifications above are collected from public sources and represent ideal conditions. Actual performance may vary depending on usage scenarios and system load.

Great For

Everyday image tasks where speed and iteration matter more than 4K pixels

Everyday Photo Editing

Remove objects, clean backgrounds, relight scenes, or retouch details — all by typing what you want

E-commerce & Product Shots

Generate product variations, swap backgrounds, and produce listing-ready images at scale

Social Media Content

Spin up feed posts, story backgrounds, and thumbnails in seconds with a 3-credit cost per image

Character & Style Consistency

Use multi-image fusion to keep a character, pet, or mascot recognizable across a short series

Rapid Concepting

Explore dozens of variations before committing — ideal for moodboards and first-round exploration

Creator & Blog Illustrations

Produce header images and inline visuals that match your post's tone, with correct aspect ratios

Nano Banana vs.

Base Model

Gemini 2.5 Flash Image

Gemini 3 Pro Image

Best For

Speed, volume, everyday edits

Studio quality, posters, text-heavy work

Reference Images

Up to 3 recommended

Up to 8 reference images

Character Consistency

Good for short series

Up to 5 distinct characters

Output Resolution

Native 1K default

Native 2K, 4K upscale available

Text Rendering

Everyday accuracy

Industry-leading, multilingual

Price per Image

Flat 3 credits

3 / 5 / 10 credits by resolution

Nano Banana (Standard)
vs
Nano Banana Pro

Frequently Asked Questions

Find answers to common questions about this model

Nano Banana is Google DeepMind's public nickname for the Gemini 2.5 Flash Image model. The name started as an internal codename during anonymous testing on LMArena in August 2025 — it went viral, and Google kept it as the official product name. It launched in Preview on August 26, 2025 and became generally available on October 2, 2025.

They are two different Google models. Nano Banana (this page) is based on Gemini 2.5 Flash Image — optimized for speed and low cost. Nano Banana Pro is based on Gemini 3 Pro Image, released in November 2025 — optimized for 4K output, multilingual text rendering, and strict multi-character consistency. Use the standard model for everyday volume work; upgrade to Pro when you need poster-grade output or precise typography.

Our platform accepts up to 10 reference files per request (JPEG, PNG, or WebP, 10 MB each). Google's official guidance is that three references or fewer produce the best fidelity — beyond that, the model has to compromise between inputs. For subject consistency, keep references tight and closely related.

Ten official ratios: 1:1, 9:16, 16:9, 3:4, 4:3, 3:2, 2:3, 5:4, 4:5, and 21:9. An Auto option is also available — the model picks the ratio that best fits your prompt. All ratios arrive at native resolution, no cropping.

Nano Banana runs on the Flash tier, so most images come back in under 10 seconds. Pricing is a flat 3 credits per image for both text-to-image and image-to-image — there's no resolution surcharge, unlike Pro.

Yes. Images generated with your account can be used for commercial purposes under our terms of service. All outputs include Google's SynthID invisible watermark, which makes AI provenance verifiable without affecting image quality. Many platforms and publishers now require or prefer watermarked AI content.

Pick Nano Banana when you need speed, volume, cheap iterations, everyday photo edits, or quick first drafts. Pick Pro when the final asset needs crisp typography (posters, thumbnails with text), 4K resolution, or strict multi-character consistency across many frames.

Yes. The same model powers text-to-image and image-to-image — no separate editor. Send a prompt alone to generate from scratch, or send a prompt plus reference images to edit, extend, or blend. The pricing is identical for either mode.

Nano Banana

Ready to Try Nano Banana?

Google DeepMind's everyday image model — fast, affordable, and as good at editing as it is at generating.

Join thousands of creators using Nano Banana