Google 推出的Nano Banana Pro指南，介绍了专业图像生成模型Nano Banana Pro的核心功能和应用技巧。文章重点强调模型在生成专业资产方面的突破，涵盖文本渲染、角色一致性、视觉合成、Google搜索联动、高级编辑、2D/3D转换、高分辨率输出等十大核心能力。文中详细提供各项功能的最佳实践提示词范例，指导用户如何通过自然语言指令像创意总监一样与模型交互，高效产出高质量的商业级图像内容。

Nano-Banana Pro相较于前几代产品实现了重大飞跃，从“娱乐性”图像生成升级到“功能性”专业资产制作。在文本渲染、字符一致性、视觉合成、世界知识（搜索）和高分辨率 (4K) 输出方面表现卓越。

本文内容包括：

复古信息图：

“制作一张 20 世纪 50 年代风格的复古信息图，介绍美国餐馆的历史。信息图应包含‘食物’、‘点唱机’和‘装饰’等独立部分。确保所有文字清晰易读，符合当时的风格。”

技术图纸：

“绘制一份正投影蓝图，以平面图、立面图和剖面图的形式描述该建筑物。使用专业建筑字体清晰标注‘北立面’和‘正门’。格式为16:9。”

白板总结（教学用途）：

用手绘白板图的形式总结“Transformer神经网络架构”的概念，适用于大学讲座。编码器和解码器模块使用不同颜色的马克笔，并清晰地标明“自注意力”和“前馈”。

角色一致性与病毒式缩略图

Nano-Banana Pro 最高支持14张参考图（其中6张可实现高保真度）。这使其能实现“身份锁定”功能——将特定人物或角色无缝置入新场景，且面部特征保持不变。

最佳实践建议：

• 身份锁定：明确指令：“确保人物面部特征与图1完全一致。”
• 表情与动作：在锁定身份的同时，可自由描述情绪或姿态的变化。
• 病毒式构图：一次性将主体与醒目的图形、文字结合，生成极具传播力的构图。

提示示例：

“病毒式缩略图”（标识 + 文字 + 图形）：

使用图 1 中的人物设计一个病毒式视频缩略图。面部一致性：保持人物面部特征与图 1 完全相同，但改变其表情，使其看起来兴奋和惊讶。动作：将人物置于画面左侧，手指指向画面右侧。主体：在右侧放置一张美味的牛油果吐司的高清图片。图形：添加一个醒目的黄色箭头，连接人物的手指和吐司。文字：在中间叠加醒目的流行风格文字：“3minuteFudede!”（3 分钟搞定！）。使用粗白线和阴影。背景：模糊明亮的厨房背景。高饱和度和高对比度。

“毛茸茸的朋友”情景（群体一致性）：

[输入3张不同毛绒玩具的图片]

请创作一个有趣的十页故事，讲述这三个毛茸茸的小伙伴去热带度假的故事。故事情节跌宕起伏，扣人心弦，最终以温馨的结局收尾。三个角色的服装和形象要保持一致，但他们的表情和角度在十幅图中​​要有所变化。每幅图中每个角色只能出现一次。

品牌资产创造：

[输入一张产品图片]

“请创作9张精美的时尚大片，风格应如同获奖时尚杂志大片。请以此为品牌风格参考，需在风格上进行细微调整和丰富变化，展现专业设计感。请一次创作一张，共创作九张图片。”

利用谷歌搜索进行真实性关联

Nano-Banana Pro 能调用谷歌搜索，依据实时数据、时事新闻或事实核查生成图像，减少在时效性话题上出现信息错漏。

最佳实践建议：

• 可要求它将动态数据（如天气、股市、新闻）进行可视化呈现。
• 模型在生成图像前，会先对搜索结果进行“思考”（推理），再开始创作。

提示示例：

活动可视化：

“根据当前的旅游趋势，生成一张信息图，展示2025年游览美国国家公园的最佳时间。”

高级编辑、修复和着色

模型擅长通过对话式指令完成复杂编辑，包括“局部重绘”（移除/添加对象）、“图像修复”（修复老照片）、“智能上色”（漫画/黑白照片）及“风格转换”。

最佳实践：

• 语义化指令：无需手动涂抹蒙版，自然描述修改需求即可。
• 物理逻辑理解：可提出复杂指令如“将玻璃杯填满液体”，检验其物理模拟生成能力。

示例提示：

物体移除与补全：

“从这张照片的背景中移除游客，并用与周围环境相匹配的合理纹理（鹅卵石和店面）填充空间。”

漫画/漫画上色：

[输入黑白漫画分镜]

“为这幅漫画分镜上色。使用鲜艳的动漫风格配色方案。确保能量光束的照明效果呈现霓虹蓝色，并且角色的服装颜色与其官方配色一致。”

本地化（文本翻译+文化适应）：

[输入一张伦敦公交车站广告的图片]

“将此概念本地化到东京场景，包括将标语翻译成日语。将背景更改为夜晚熙熙攘攘的涩谷街道。”

照明/季节控制：

[输入一张夏季房屋的图片]

“将此场景转换为冬季。保持房屋建筑结构完全相同，但在屋顶和院子里添加积雪，并将照明更改为寒冷阴沉的午后。”

维度转换（2D ↔ 3D）

一项强大的新功能，能将二维平面图转化为三维可视化模型，反过来也可以。对于室内设计师、建筑师乃至 meme 创作者来说，都十分理想。

提示示例：

2D平面图转3D室内设计效果图：

根据上传的2D平面图，生成一张专业的室内设计效果图。布局：采用拼贴画形式，顶部为一张主图（客厅广角视图），下方为三张小图（主卧、家庭办公室和3D俯视图）。风格：所有图片均采用现代简约风格，搭配温暖的橡木地板和米白色墙面。质量：照片级渲染，柔和的自然光。

2D 到 3D 表情包转换：

“将‘一切都好’狗狗表情包转换成逼真的 3D 渲染图。保持构图不变，但让狗狗看起来像毛绒玩具，火焰看起来像真实的火焰。”

高分辨率和纹理增强

Nano-Banana Pro 原生支持1K至4K的图像生成。此特性对于表现细腻纹理或制作大尺寸印刷品尤为实用。

最佳实践建议：

• 高分辨率请求：若您的API或界面允许，请明确要求生成2K或4K的高分辨率图像。
• 高保真细节描述：在提示词中描述诸如细微瑕疵、复杂表面纹理等高保真细节。

4K纹理生成：

“利用原生高保真输出，打造令人叹为观止、充满氛围的苔藓森林地面环境。掌控复杂的光照效果和细腻的纹理，确保每一根苔藓和每一束光线都以像素级分辨率渲染，满足4K壁纸的需求。”

复杂逻辑（思考模式）：

“制作一张超逼真的高级芝士汉堡信息图，将其拆解，展现烤过的奶油蛋卷面包的质地、肉饼煎至焦香的外皮以及闪闪发光的融化芝士。为每一层标注其风味特征。”

思考与推理能力

Nano-Banana Pro 默认采用“思考”模式，模型会生成一些中间的思考图像（不计费），以便在渲染最终输出之前优化构图。有助于进行数据分析和解决视觉问题。

示例提示：

解方程：

在白板上用 C 语言求解方程 log_{x^2+1}(x^4-1)=2。请清晰地写出解题步骤。

视觉推理：

“分析这张房间图片，生成一张‘之前’的图片，展示房间在施工期间可能的样子，包括框架和未完成的石膏板。”

一次性故事板和概念设计

您无需借助分格模板，即可直接生成序列图像或故事板，确保在单次会话中构建出连贯的叙事流。此功能广泛应用于制作“电影概念艺术”（例如，发布即将上映影片的虚假泄露图）。

提示示例：

请创作一个引人入胜的九部分故事，包含九张图片，故事中需出现一位女性和一位男性，他们正在拍摄一部屡获殊荣的豪华行李箱广告。故事应有跌宕起伏的情感，最后以一位女性手持品牌标识的优雅照片结尾。男女主角的身份和着装必须保持一致，但可以从不同的角度和距离拍摄。请逐一生成图片。请确保每张图片均为16:9横向格式。

结构控制与布局引导

参考图的应用不仅限于角色或待编辑对象。您可以通过它们严格控制最终图像的构图与布局。对于需要将草稿、线框图或特定网格转化为精美资产的设计师而言，无疑是一项颠覆性功能。

最佳实践建议：

• 草案与草图：上传手绘草图，精确指定文本和物体的位置。
• 线框图：使用现有布局或线框图的截图，生成高保真UI效果图。
• 网格：利用网格图片，驱动模型生成专为瓦片式游戏或LED显示屏设计的图像资产。

提示示例：

从草图到最终广告：

“根据此草图为[产品]创作一个广告。”

根据线框图创建 UI 模型：

“按照以下准则为[产品]创建模型。”

像素艺术与LED显示屏：

“请绘制一个独角兽像素画，使其完美契合这幅 64×64 的网格图像。使用高对比度的颜色。”

（提示：开发人员随后可以通过编程方式提取每个单元格的中心颜色，以驱动连接的 64×64 LED 点阵显示屏）。

精灵图示例：

“一位女性在无人机上做后空翻的精灵图，3×3网格，逐帧动画序列，正方形宽高比。请完全按照附件参考图像的结构进行绘制。”

（提示：您可以提取每个单元格并制作成 GIF 动画）

进阶指南

既然已掌握了提示词的基本要领，接下来可以开始构建：

• 在界面中实验：Google AI Studio 是测试提示词和参数的最快捷方式。
• 查看精彩应用：在应用画廊中，体验由 Nano-Banana 驱动的酷炫应用。
• 将创意转化为应用：在 AI Studio Build 中，将您最成功的提示词轻松转化为可分享给朋友的应用。
• 构建应用程序：准备编写代码？请查阅开发者指南或 Gemini API 示例库，获取指南和代码片段。
• 技术深度探索：阅读完整的 Gemini API 文档，了解关于速率限制、定价和集成等详细信息。

 英文原文

Nano-Banana Pro is a significant leap forward from previous generation models, moving from “fun” image generation to “functional” professional asset production. It excels in text rendering, character consistency, visual synthesis, world knowledge (Search), and high-resolution (4K) output.

Following the developer guide on how to get started with AI Studio and the API, this guide covers the core capabilities and how to prompt them effectively.

By Guillaume Vernade, Gemini Developer Advocate, Google DeepMind

Here’s what you’ll find in this article:

1. The Golden Rules of Prompting
2. Text Rendering, Infographics & Visual Synthesis
3. Character Consistency & Viral Thumbnails
4. Grounding with Google Search
5. Advanced Editing, Restoration & Colorization
6. Dimensional Translation (2D ↔ 3D)
7. High-Resolution & Textures
8. Thinking & Reasoning
9. One-Shot Storyboarding & Concept Art
10. Structural Control & Layout Guidance
11. What’s Next?

🛑 Section 0: The Golden Rules of Prompting

Nano-Banana Pro is a “Thinking” model. It doesn’t just match keywords; it understands intent, physics, and composition. To get the best results, stop using “tag soups” (e.g., dog, park, 4k, realistic) and start acting like a Creative Director.

Edit, Don’t Re-roll

The model is exceptionally good at understanding conversational edits. If an image is 80% correct, do not generate a new one from scratch. Instead, simply ask for the specific change you need.

Example: “That’s great, but change the lighting to sunset and make the text neon blue.”

Use Natural Language & Full Sentences

Talk to the model as if you were briefing a human artist. Use proper grammar and descriptive adjectives.

❌ Bad: “Cool car, neon, city, night, 8k.”

✅ Good: “A cinematic wide shot of a futuristic sports car speeding through a rainy Tokyo street at night. The neon signs reflect off the wet pavement and the car’s metallic chassis.”

Be Specific and Descriptive

Vague prompts yield generic results. Define the subject, the setting, the lighting, and the mood.

Subject: Instead of “a woman,” say “a sophisticated elderly woman wearing a vintage chanel-style suit.”

Materiality: Describe textures. “Matte finish,” “brushed steel,” “soft velvet,” “crumpled paper.”

Provide Context (The “Why” or “For whom”)

Because the model “thinks,” giving it context helps it make logical artistic decisions.

Example: “Create an image of a sandwich for a Brazilian high-end gourmet cookbook.” (The model will infer professional plating, shallow depth of field, and perfect lighting).

Text Rendering, Infographics & Visual Synthesis

Nano-Banana Pro has SOTA capabilities for rendering legible, stylized text and synthesizing complex information into visual formats.

Best Practices:

• Compression: Ask the model to “compress” dense text or PDFs into visual aids.
• Style: Specify if you want a “polished editorial,” a “technical diagram,” or a “hand-drawn whiteboard” look.
• Quotes: Clearly specify the text you want in quotes.

Example Prompts:

Earnings Report Infographic (Data Ingestion):

[Input PDF of Google’s latest earnings report]

“Generate a clean, modern infographic summarizing the key financial highlights from this earnings report. Include charts for ‘Revenue Growth’ and ‘Net Income’, and highlight the CEO’s key quote in a stylized pull-quote box.”

Retro Infographic:

“Make a retro, 1950s-style infographic about the history of the American diner. Include distinct sections for ‘The Food,’ ‘The Jukebox,’ and ‘The Decor.’ Ensure all text is legible and stylized to match the period.”

Technical Diagram:

“Create an orthographic blueprint that describes this building in plan, elevation, and section. Label the ‘North Elevation’ and ‘Main Entrance’ clearly in technical architectural font. Format 16:9.”

Whiteboard Summary (Educational):

“Summarize the concept of ‘Transformer Neural Network Architecture’ as a hand-drawn whiteboard diagram suitable for a university lecture. Use different colored markers for the Encoder and Decoder blocks, and include legible labels for ‘Self-Attention’ and ‘Feed Forward’.”

Character Consistency & Viral Thumbnails

Nano-Banana Pro supports up to 14 reference images (6 with high fidelity). This allows for “Identity Locking”—placing a specific person or character into new scenarios without facial distortion.

Best Practices:

• Identity Locking: Explicitly state: “Keep the person’s facial features exactly the same as Image 1.”
• Expression/Action: Describe the change in emotion or pose while maintaining the identity.
• Viral Composition: Combine subjects with bold graphics and text in a single pass.

Example Prompts:

The “Viral Thumbnail” (Identity + Text + Graphics):

“Design a viral video thumbnail using the person from Image 1. Face Consistency: Keep the person’s facial features exactly the same as Image 1, but change their expression to look excited and surprised. Action: Pose the person on the left side, pointing their finger towards the right side of the frame. Subject: On the right side, place a high-quality image of a delicious avocado toast. Graphics: Add a bold yellow arrow connecting the person’s finger to the toast. Text: Overlay massive, pop-style text in the middle: ‘3分钟搞定!’ (Done in 3 mins!). Use a thick white outline and drop shadow. Background: A blurred, bright kitchen background. High saturation and contrast.”

The “Fluffy Friends” Scenario (Group Consistency):

[Input 3 images of different plush creatures]

“Create a funny 10-part story with these 3 fluffy friends going on a tropical vacation. The story is thrilling throughout with emotional highs and lows and ends in a happy moment. Keep the attire and identity consistent for all 3 characters, but their expressions and angles should vary throughout all 10 images. Make sure to only have one of each character in each image.”

Brand Asset Generation:

[Input 1 image of a product]

“Create 9 stunning fashion shots as if they’re from an award-winning fashion editorial. Use this reference as the brand style but add nuance and variety to the range so they convey a professional design touch. Please generate nine images, one at a time.”

Grounding with Google Search

Nano-Banana Pro uses Google Search to generate imagery based on real-time data, current events, or factual verification, reducing hallucinations on timely topics.

Best Practices:

• Ask for visualizations of dynamic data (weather, stocks, news).
• The model will “Think” (reason) about the search results before generating the image.

Example Prompts:

Event Visualization:

“Generate an infographic of the best times to visit the U.S. National Parks in 2025 based on current travel trends.”

Advanced Editing, Restoration & Colorization

The model excels at complex edits via conversational prompting. This includes “In-painting” (removing/adding objects), “Restoration” (fixing old photos), “Colorization” (Manga/B&W photos), and “Style Swapping.”

Best Practices:

• Semantic Instructions: You do not need to manually mask; simply tell the model what to change naturally.
• Physics Understanding: You can ask for complex changes like “fill this glass with liquid” to test physics generation.

Example Prompts:

Object Removal & In-painting:

“Remove the tourists from the background of this photo and fill the space with logical textures (cobblestones and storefronts) that match the surrounding environment.”

Manga/Comic Colorization:

[Input black and white manga panel]

“Colorize this manga panel. Use a vibrant anime style palette. Ensure the lighting effects on the energy beams are glowing neon blue and the character’s outfit is consistent with their official colors.”

Localization (Text Translation + Cultural Adaptation):

[Input image of a London bus stop ad]

“Take this concept and localize it to a Tokyo setting, including translating the tagline into Japanese. Change the background to a bustling Shibuya street at night.”

Lighting/Seasonal Control:

[Input image of a house in summer]

“Turn this scene into winter time. Keep the house architecture exactly the same, but add snow to the roof and yard, and change the lighting to a cold, overcast afternoon.”

Dimensional Translation (2D ↔ 3D)

A powerful new capability is translating 2D schematics into 3D visualizations, or vice versa. This is ideal for interior designers, architects, and meme creators.

Example Prompts:

2D Floor Plan to 3D Interior Design Board:

“Based on the uploaded 2D floor plan, generate a professional interior design presentation board in a single image. Layout: A collage with one large main image at the top (wide-angle perspective of the living area), and three smaller images below (Master Bedroom, Home Office, and a 3D top-down floor plan). Style: Apply a Modern Minimalist style with warm oak wood flooring and off-white walls across ALL images. Quality: Photorealistic rendering, soft natural lighting.”

2D to 3D Meme Conversion:

“Turn the ‘This is Fine’ dog meme into a photorealistic 3D render. Keep the composition identical but make the dog look like a plush toy and the fire look like realistic flames.”

High-Resolution & Textures

Nano-Banana Pro supports native 1K to 4K image generation. This is particularly useful for detailed textures or large-format prints.

Best Practices:

• Explicitly request high resolutions (2K or 4K) if your API/Interface allows.
• Describe high-fidelity details (imperfections, surface textures).

4K Texture Generation:

“Harness native high-fidelity output to craft a breathtaking, atmospheric environment of a mossy forest floor. Command complex lighting effects and delicate textures, ensuring every strand of moss and beam of light is rendered in pixel-perfect resolution suitable for a 4K wallpaper.”

Complex Logic (Thinking Mode):

“Create a hyper-realistic infographic of a gourmet cheeseburger, deconstructed to show the texture of the toasted brioche bun, the seared crust of the patty, and the glistening melt of the cheese. Label each layer with its flavor profile.”

Thinking & Reasoning

Nano-Banana Pro defaults to a “Thinking” process where it generates interim thought images (not charged) to refine composition before rendering the final output. This allows for data analysis and solving visual problems.

Example Prompts:

Solve Equations:

“Solve log_{x^2+1}(x^4-1)=2 in C on a white board. Show the steps clearly.”

Visual Reasoning:

“Analyze this image of a room and generate a ‘before’ image that shows what the room might have looked like during construction, showing the framing and unfinished drywall.”

One-Shot Storyboarding & Concept Art

You can generate sequential art or storyboards without a grid, ensuring a cohesive narrative flow in a single session. This is also popular for “Movie Concept Art” (e.g., fake leaks of upcoming films).

Example Prompt:

“Create an addictively intriguing 9-part story with 9 images featuring a woman and man in an award-winning luxury luggage commercial. The story should have emotional highs and lows, ending on an elegant shot of the woman with the logo. The identity of the woman and man and their attire must stay consistent throughout but they can and should be seen from different angles and distances. Please generate images one at a time. Make sure every image is in a 16:9 landscape format.”

Structural Control & Layout Guidance

Input images aren’t limited to character references or subjects to edit. You can use them to strictly control the composition and layout of the final output. This is a game-changer for designers who need to turn a napkin sketch, a wireframe, or a specific grid layout into a polished asset.

Best Practices:

• Drafts & Sketches: Upload a hand-drawn sketch to define exactly where the text and object should sit.
• Wireframes: Use screenshots of existing layouts or wireframes to generate high-fidelity UI mockups.
• Grids: Use grid images to force the model to generate assets for tile-based games or LED displays.

Sketch to Final Ad:

“Create a ad for a [product] following this sketch.”

UI Mockup from Wireframe:

“Create a mock-up for a [product] following these guidelines.”

Pixel Art & LED Displays:

“Generate a pixel art sprite of a unicorn that fits perfectly into this 64×64 grid image. Use high contrast colors.”

(Tip: Developers can then programmatically extract the center color of each cell to drive a connected 64×64 LED matrix display).

Sprites:

“Sprite sheet of a woman doing a backflip on a drone, 3×3 grid, sequence, frame by frame animation, square aspect ratio. Follow the structure of the attached reference image exactly..”

(Tip: You can then extract each cell and make a gif)

What’s Next?

Now that you have mastered the basics of prompting, here is how you can start building:

• Experiment in the UI: Google AI Studio is the fastest way to test prompts and parameters.
• Check really cool Nano-banana powered app in the App Gallery.
• Vibe-code you dream app: Transform you best prompt into an app that you can easily share with your friends in AI Studio Build.
• Build Applications: Ready to code? Check out the developer guide or the Gemini API Cookbook for guides and code snippets.
• Technical Deep Dive: Read the full Gemini API Documentation for details on rate limits, pricing, and integration.