Nano Banana 2
热门Nano Banana 2 is the Google image generation and editing model family built on Gemini 3.1 Flash, pairing high fidelity native image creation with multi turn conversational edits, consistent characters, and volume friendly pricing for creative and production workflows.
概述
Nano Banana 2 is the current generation of the Google image model family that went viral in 2025, now running on the Gemini 3.1 Flash architecture under the endpoint gemini-3.1-flash-image and marked stable in the Gemini API. It generates images from text, then refines them through multi turn conversational editing, keeping characters, wardrobe, and scene layout consistent across iterations, which makes it effective for ad variants, character sheets, and product visuals. Output scales from 512px drafts to 4K masters, and the model is tuned for high volume interactive use with Flash-class latency. API image output is token based at $60 per 1M image tokens, which works out to about $0.045 for a 512px image, $0.067 at 1024 by 1024, and up to roughly $0.151 at 4K, and every image carries a SynthID provenance watermark. The family spans three tiers: Nano Banana 2 for the balance of quality and speed, Nano Banana 2 Lite for ultra low latency and cost, and Nano Banana Pro for state of the art contextual rendering on the gemini-3-pro-image endpoint. Access runs through Google AI Studio for free experimentation, the Gemini API and Vertex AI for developers, and the Gemini app plus Google AI subscription plans for consumers.
核心功能
优点
- +Best in class multi turn image editing
- +Character consistency across iterations
- +Volume friendly token pricing from about $0.045 per image
缺点
- -Best editing depth requires Gemini API familiarity
- -Photoreal text rendering still imperfect on complex layouts
适用场景
集成与兼容性
Frequently Asked Questions
What is Nano Banana 2?
Nano Banana 2 is the 2026 generation of the Google image generation and editing model family, built on Gemini 3.1 Flash. It creates images from text, refines them with multi turn edits, and outputs from 512px up to 4K, available free in Google AI Studio and through the Gemini API.
How much does Nano Banana 2 cost?
Experimentation in Google AI Studio is free, and API image output is priced at $60 per 1M tokens, about $0.045 per 512px image, $0.067 at 1024px, and up to $0.151 for a 4K image. Consumer access is included with Google AI subscription plans.
How does Nano Banana 2 compare to Midjourney?
Nano Banana 2 leads on conversational multi turn editing and character consistency, since you refine the same image by describing changes. Midjourney still attracts many artists for its aesthetic presets, while Nano Banana 2 wins on speed, volume pricing, and integration with the Gemini ecosystem.
Can Nano Banana 2 keep characters consistent across images?
Yes, consistent characters and scenes are a signature strength. The model holds identity, wardrobe, and style across sequential edits and generations, which suits storyboards, ad variants, and character sheets.
What is the difference between Nano Banana 2, 2 Lite, and Pro?
Nano Banana 2 is the balanced tier for quality and speed, Nano Banana 2 Lite trades some fidelity for ultra low latency and cost on high volume jobs, and Nano Banana Pro is the flagship for the most contextual, detailed renders on the gemini-3-pro-image endpoint.
How do I use Nano Banana 2 in the Gemini app?
Open the Gemini app or website, start a chat, and describe the image you want, then refine it through follow up messages. Google AI subscription plans include image generation with higher limits, and the same model powers editing when you upload a photo and describe the change you want.
What resolution can Nano Banana 2 output?
Output scales from 512 pixel drafts up to 4K masters. API pricing is token based at $60 per 1M image tokens, which works out to about $0.045 for a 512 pixel image, $0.067 at 1024 by 1024, and roughly $0.151 at 4K, so cost scales with the resolution you actually need.
Can it add or remove objects in an existing photo?
Yes. Multi turn conversational editing is the core strength: upload a photo, then ask for changes such as removing a background object, swapping a product color, or adding an element, and the model keeps the rest of the scene stable. Each edit becomes a new turn you can redo by regenerating.
Is Nano Banana 2 good enough for product photos?
For e-commerce and ad variants it is one of the strongest options because it keeps products, wardrobe, and scene layout consistent across a set. At roughly $0.045 to $0.067 per image on the API, a batch of product shots costs cents rather than a studio budget. Every output carries a SynthID watermark.
Is there a free way to try it before paying?
Yes. Google AI Studio offers free experimentation with the model, and the Gemini app includes image generation on free and paid consumer plans within usage limits. For production use, the Gemini API and Vertex AI bill per image, starting at about $0.045 for the smallest size.
评分明细
支持的语言
数据隐私与安全
Processed under Google terms with SynthID provenance watermarking on outputs