Nano Banana
参数配置
生成结果
样例展示
A cozy coffee shop interior on a rainy morning, steam rising from a ceramic mug, warm amber lighting, bokeh background
规格参数
定价
credits per image
~3 credits for a 1024x1024 image
Nano Banana is the official nickname given by the Google Gemini team to Gemini 2.5 Flash Image — a fast, multimodal image generation model built on Google's Gemini 2.5 Flash architecture. Its defining capability is conversational image editing: you can generate an image, then refine it through a natural language back-and-forth without starting over. The model understands context accumulated across the conversation, so each edit builds on previous results rather than replacing them.
能力说明
Nano Banana excels in three distinct areas. First, multi-turn iterative editing: describe an initial scene, then adjust lighting, composition, colour palette, or specific objects through follow-up natural language instructions — the model applies changes surgically while keeping the rest of the image intact. Second, cross-image character and scene consistency: within a conversation session, the model can reproduce a defined character across different poses and backgrounds, enabling coherent storyboards and sequential visual narratives. Third, generation speed: being built on the Flash-tier Gemini architecture, it produces images quickly, making tight creative loops feel genuinely fluid rather than like waiting for a render.
最佳实践
To get the most from Nano Banana, treat it as a creative conversation partner rather than a one-shot prompt engine. Start with a clear initial description of the scene, character, or mood, then refine iteratively: "now make the background a forest at dusk" or "give her a blue scarf". When consistency matters across multiple images, establish the character or environment in the first turn and reference it explicitly in subsequent requests: "same character, now in a rainy city street". Because the model retains conversational context, each successive instruction can be concise — you only need to describe what changes, not repeat everything.
常见问题
What is Nano Banana and which underlying model does it use?
Nano Banana is the official playful nickname for Google Gemini 2.5 Flash Image, Google's fast multimodal image generation model. It is not a separate model — the name "Nano Banana" was coined by the Gemini team as a friendly handle for the image-generation capability built into Gemini 2.5 Flash.
How does conversational image editing work with Nano Banana?
Unlike traditional text-to-image models that treat each request independently, Nano Banana supports multi-turn dialogue. You can generate an initial image, then refine it through natural follow-up instructions: "make the sky more dramatic", "add a red umbrella in the foreground", or "switch to evening lighting" — and the model applies each edit while preserving the rest of the scene. This conversational loop dramatically speeds up the creative iteration process.
What is character and scene consistency in Nano Banana?
Nano Banana can maintain the identity of a character or the visual style of a scene across multiple generated images within the same conversation. If you define a character — appearance, clothing, expression — the model keeps that character recognisably consistent when you ask for new poses, environments, or situations. This makes it particularly well-suited for storyboarding, comic strips, and multi-scene social media content.
相关模型
FLUX Pro
Black Forest Labs
Generate photorealistic images at up to 2048x2048 with FLUX Pro by Black Forest Labs. Industry-leading prompt adherence and detail fidelity from the original Stable Diffusion research team. Commercially licensed output for professional creative workflows.
Ideogram v3
Ideogram
Create logos, posters, and typographic designs with perfect text rendering using Ideogram v3. The only image AI that consistently generates legible, stylised text inside images — ideal for brand identity, marketing, and editorial design.
FLUX 2
Black Forest Labs
FLUX 2 by Black Forest Labs is the next generation of the acclaimed FLUX architecture, promising native 4K image generation, faster inference, and even stronger prompt adherence than FLUX Pro. Join the Genace waitlist to access it at launch.