Nano Banana 2

Google DeepMind's latest image generation model. Combines the quality and world knowledge of Nano Banana Pro with Flash-tier speed, at roughly half the API cost.

What it does

Nano Banana 2 (Gemini 3.1 Flash Image) generates and edits photorealistic images from text prompts. It ranks #1 on both the Artificial Analysis Image Arena and LM Arena for text-to-image. It ties for #1 in single-image editing with ChatGPT-Image-Latest.

The model pulls from Gemini's knowledge base and real-time web search to render specific subjects accurately. Ask for a real landmark or product and it references current visual data, not training data alone.

Key specs

• Resolution: 512px to 4K

• Subject consistency: Up to 5 characters and 14 objects per workflow

• Text rendering: Accurate in-image text with multi-language translation

• Aspect ratios: Multiple formats (vertical, widescreen, square)

• Web grounding: Real-time search for accurate subject rendering

• Provenance: SynthID watermarking + C2PA Content Credentials

Good for

• Product imagery: Consistent product shots across settings and styles

• Marketing mockups: Posters, ads, and social posts with accurate text

• Storyboarding: Maintain character consistency across scenes

• Style transfer: Apply reference photo aesthetics to your subject

• Infographics: Diagrams, data visualizations, and labeled visuals

• Localization: Translate text within images across languages

What's improved over the original Nano Banana

• Pro-level quality at Flash speed

• Subject consistency (5 characters, 14 objects)

• 4K resolution support (up from 1024x1024)

• Precision text rendering and translation

• Web search grounding for real-world accuracy

• Stronger instruction following for complex prompts

About

Made by Google DeepMind. Released February 26, 2026. Successor to Nano Banana (August 2025) and Nano Banana Pro (November 2025).

Frequently Asked Questions

What is the difference between Nano Banana 2 and Nano Banana Pro? Nano Banana 2 matches or beats Pro on most benchmarks while running at Flash speed. It costs roughly half the price per image. Pro remains available for tasks needing maximum factual accuracy through the Gemini app's regeneration menu.

How does Nano Banana 2 compare to ChatGPT image generation? Nano Banana 2 ranks #1 in text-to-image on both the Artificial Analysis Image Arena and LM Arena. It ties with ChatGPT-Image-Latest for #1 in single-image editing. It's faster (Flash-tier) and cheaper per image.

Can Nano Banana 2 generate text inside images? Yes. The model renders accurate, legible text for logos, posters, greeting cards, and marketing materials. You can also translate and localize text within images across multiple languages.

What is the maximum resolution for Nano Banana 2? 4K. The model supports outputs from 512px up to 4K across multiple aspect ratios including vertical, widescreen, and square formats.

Can I use Nano Banana 2 images commercially? Commercial usage depends on Google's Gemini API terms of service. All generated images include SynthID watermarking and C2PA Content Credentials for provenance tracking. Check Google's current policies for specifics.

How do I run Nano Banana 2 on Floyo? Find a workflow that includes the Gemini API node, then click Run. No API keys to configure, no installation. Floyo handles the connection. You can combine Nano Banana 2 with open-source models and custom nodes in the same pipeline.