★ Bookmark | Just speak human language, guiding you through AIGC.Since the release of Nano Banana Pro, it seems that all AI players online are using it to “create magic”.Like everyone else, I have looked at various tutorials online and copied and pasted prompts.However, after trying many times, you will find that these unofficial guides often only explain the surface without understanding the underlying principles; they are either too one-sided or fail in different scenarios.This feeling of “blind box” is quite frustrating.Fortunately, Google has released an official guide, which provides a relatively systematic introduction.
Image: Official NBP guide released by GoogleI have retested and verified each function. I have rewritten (almost) all the cases to better suit the needs of Chinese users.
This article is the “fruit of my labor” that I organized after spending an entire night, hoping to help you connect those fragmented pieces of knowledge all at once.(I have bundled all the prompts in an attachment, and the method to obtain it is at the end of the article.)
01｜Even a financial report can be turned into a work of art
Creating infographics used to be a designer’s nightmare.
Now? Just throw a bunch of complex text at NBP, and it can organize it clearly in no time.
Let’s look at a few examples.
Retro-style infographic: I asked it to create an infographic about the history of Hong Kong-style tea restaurants, with a touch of the 1980s.
Prompt:A retro-style infographic from the 1980s about the history of Hong Kong-style tea restaurants. Please include three independent sections: ‘classic dishes’, ‘nostalgic tiles’, and ‘neon signs’. Ensure all text is clear and readable, and use a font design that fits the aesthetics of that era.Financial report turns into a poster: The weekly report savior for workers is here.No more drawing in Excel; directly generate this modern-style chart that will leave your boss stunned.
Prompt:Generate a clean, modern infographic in Chinese summarizing the key financial highlights from this financial report. Include charts for ‘revenue growth’ and ‘net profit’, and highlight the CEO’s key quotes in a stylized quote box. Do not include citation labels, use Google’s color scheme.
Technical drawings:If you are in engineering or architectural design, creating such orthographic blueprints used to take ages in CAD.
Now NBP can directly generate these professional-looking drawings.
Prompt:Create an orthographic blueprint that describes this building through floor plans, elevations, and sections. Clearly label ‘North Elevation’ and ‘Main Entrance’ using technical architectural fonts. Format should be 16:9.
Even… hand-drawn whiteboards
Even complex concepts like “mobile payment principles” can be mimicked as sketches drawn by a university professor on a whiteboard, complete with the strokes of a marker.
Prompt:Summarize the concept of ‘how the Chinese mobile payment ecosystem works’ (like Alipay or WeChat Pay) into a hand-drawn whiteboard diagram suitable for a university lecture. Use different colored markers to distinguish between ‘user/merchant front-end’ and ‘bank/settlement back-end’ sections, and include clear labels for ‘scan interaction’ and ‘encrypted settlement’.
For this type of diagram, Google provides three key tips:
-
Information condensation: Whether it’s a long article or a PDF, throw it at it and tell it to “extract the key points”.
-
Style setting: Do you want a “magazine style”, “technical blueprint”, or “hand-drawn whiteboard”? Be clear about it.
-
Specify content: For example, in a financial report, if you only care about “how much was earned”, instruct it to only draw the profit section.
02｜Finally! Characters no longer “swap faces”.
What is the biggest pain point in AI drawing? Character consistency.The previous image might have a heart-shaped face, and the next one turns into a square face, making it impossible to create a coherent story.NBP has truly evolved this time.It supports up to 14 reference images! This is like taking the actor’s makeup photos and directly applying them to the AI face, telling it: “This person, lock it in, don’t change.”For example, I want to create a viral video cover for this character:
Prompt:Using the character from image 1, design a viral video cover. Character consistency: Keep the character’s facial features exactly the same as in image 1, but change the expression to one of extreme shock and disbelief, with an open mouth. Action: Position the character on the right side of the image, with both hands on their head, looking towards the left side of the image. Theme: On the left side of the image, place a high-definition close-up of the latest domestic flagship phone, with the screen lit up showing benchmark scores. Graphic elements: Add explosion-like radiating effect lines around the phone, and use a huge red arrow pointing to the phone screen. Text: Overlay large, variety show-style text in the center of the image: ‘This performance is insane?!’. Use bold white outlines and strong black shadows. Background: A blurred, neon-lit esports room background. High saturation and high contrast.
Can anyone guess who this cover character is? I feel the character’s image is well maintained.
Not just characters, you can use a logo to design a bunch of materials while keeping that logo consistent everywhere.
Here, I will use my logo as an example.
Prompt:[Upload logo image] Use this logo to create various types of branding materials for a bookstore named “AI Fan’er”.
NBP can also maintain consistency for multiple objects, up to 5.
This consistency for groups is essential when making story films.
Prompt:[Input 3 images of different plush creatures] “Create an interesting story in 10 parts about the experiences of these 3 plush friends on a tropical vacation. The story should be thrilling and emotional, ending with a happy moment. Keep the costumes and identities of all 3 characters consistent, but their expressions and angles should vary in all 10 images. Ensure each character appears only once in each image.
03｜It is connected to the internet!
This trick is unknown to many. Remember to add a line in the prompt: “Combine with Google Search”.Unlike other models that are oblivious to the outside world, NBP is connected to Google Search.This means it knows today’s stock market, tomorrow’s weather, and even the latest hot news.For example, I want to enjoy cherry blossoms but don’t know the best time to go?
Prompt:Combine with Google Search to generate a visual infographic showing the best viewing time windows for major cherry blossom viewing spots in China in 2025 (such as Wuhan University, Wuxi Yuantouzhu) based on current climate predictions and tourism big data trends.
For example, I want to know the distribution of the eight major cuisines in China.
Prompt:Combine with Google Search to generate a food culture infographic based on a map of China, visualizing the ‘geographical distribution of the eight major cuisines in China’. Please clearly label the cuisine schools in the corresponding areas of the map (e.g., label ‘Sichuan cuisine’ in the Sichuan area, ‘Cantonese cuisine’ in the Guangdong area), and include a hand-drawn illustration of the most representative signature dish for each cuisine section.
Being able to query online at any time is truly a great innovation in text-to-image generation.
04｜”One-sentence photo editing”, goodbye Photoshop
This is a feature that “clients love”.
“Remove the people in the background.” “Color this painting.” “Change daytime to nighttime.”
These requests used to take a long time to modify, but now it’s just a matter of one sentence.
Remove passersby
Prompt:Remove all tourists from the background of this photo. Use textures that match the ancient blue brick stone ground and red wall architecture of the Forbidden City to logically fill in the empty areas, making the entire square look completely empty and realistic. Keep the character’s image and posture unchanged.
Color the comic
Prompt:Color this comic.
Localize the scene
Prompt:Localize this image to a Shanghai scene, including translating the slogans into Chinese. Change the background to a busy Shanghai street at night.
Change the seasons
Prompt:Transform this scene into winter. Keep the house structure unchanged, but add snow on the roof and yard, and change the lighting to a cold, gloomy afternoon.
05｜2D to 3D, designers rejoice
This feature is a dimensionality reduction strike.
Throw your flat sketches in, and it can directly generate 3D renderings; or conversely, turn 3D models into flat structure diagrams.
Interior designers, architects, and even meme creators must play with this.
Prompt:Based on the uploaded 2D floor plan, generate a single-page professional interior design proposal collage. Layout: Place a wide-angle perspective main image showcasing the living and dining area at the top, highlighting the iconic round dining table area and open kitchen island; at the bottom, display three small images side by side, detailing the main bathroom with a bathtub, a study perspective, and a 3D overhead view of the unit. The overall style should be modern minimalist, with warm oak flooring and off-white walls, ensuring all images have natural soft lighting to achieve photo-level rendering quality.
Prompt:Transform the ‘muscle Pepe stock trading’ meme image into a realistic 3D rendering. Keep the composition exactly the same as the original, but sculpt muscle Pepe into a highly realistic biological frog with moist skin and clear muscle textures; simultaneously, render the laptop screen as a real self-illuminating LCD display, and the four candles on the table with realistic wax drip textures and flickering flames, casting warm, realistic light and shadows on the wooden table.
06｜Detail fanatic? Please enable high-definition mode
If you are in printing, publishing, or like me, a “detail-oriented person”, the native support for 1K, 2K, and even 4K output from NBP is crucial.
If using a dialogue box like Gemini, remember to include keywords like “high definition”, “HD”, “4K”, and you can describe material details, such as “leather texture”, “metal scratches”.
Prompt:Utilize native high-fidelity output to create a stunning, atmospheric close-up environment of a rain-soaked ancient temple roof in Jiangnan. Command complex wet light and shadow effects and intricate aged material textures, ensuring every drop of rain reflects on the aged wooden beams, and every detail of moss in the tile gaps is presented in pixel-perfect resolution suitable for 4K wallpaper.
In addition to this texture requirement, it also supports such complex logic.
Prompt:Create a surreal infographic of Cantonese double yolk mooncakes, cut open or three-dimensional deconstructed to showcase the golden, oily crust, the delicate and dense red lotus seed paste filling, and the soft, flowing salted egg yolk at the center. Please detail each layer’s unique texture and flavor source in the infographic.
07｜It not only draws but also solves Olympiad math problems
Don’t forget, NBP combines the brain of Gemini 3 Pro. It “thinks” before drawing. This means it can handle logical problems.
For example, let it solve a logarithmic equation in the complex domain and write out the steps: (from the official case)
Prompt:Solve $log_{x^2+1}(x^4-1)=2$ in the complex domain C on a whiteboard. Clearly show the steps.
Or, show it a photo of a well-decorated living room and let it reverse engineer what the “bare shell” house looked like before renovation:
Prompt: Analyze this photo of a modern cozy living room after renovation and generate an image of its state before renovation as a “bare shell house”. Please showcase exposed cement walls, complex wiring pipes on the floor, and rough concrete textures, restoring the authentic scene from the initial construction phase.
08｜Storyboard from a movie director’s perspective
Want to draw comics or movie storyboards? No need to generate them one by one. You can ask it to draw multiple consecutive scenes in one image.For example, this story:A pair of male and female protagonists dressed in new Chinese-style clothing, depicting a high-end tea brand advertisement narrative scene returning from the hustle and bustle of the city to the tranquility of the mountains.
Prompt:See attachment to conceive an engaging 9-part visual story, presented through 9 images, themed around a pair of male and female protagonists starring in an award-winning cultural advertisement for a high-end Chinese tea brand. The storyline needs to include emotional ups and downs and twists (e.g., from the hustle and confusion of the city to the tranquil realization in the mountain tea garden), ending with an elegant freeze-frame shot of the female protagonist holding a teacup alongside the brand logo. Ensure the facial features and clothing styles of the male and female protagonists (such as new Chinese-style clothing) remain strictly consistent throughout the series, but showcase rich variations in photography angles and shot sizes (such as aerial views, medium shots, and close-up details). Please generate these nine images one by one, ensuring each is in a 16:9 horizontal format.
In fact, Gemini also supports generating multiple images at once.
09｜Control composition like a puppet master
This last trick is a big deal. In addition to controlling character appearances with images, you can also control “structure” and “layout” with images.
You can casually draw a sketch (even if it’s ugly), or give it a specified grid layout, and it will obediently fill in the picture.
Prompt:[Upload a hand-drawn image] Based on this hand-drawn wireframe structure, create a high-fidelity ‘new Chinese-style national tide tea drink’ mobile UI interface design. The overall style should adopt an elegant Ru kiln sky blue and beige tone, incorporating an ink wash texture background. The top ‘title bar’ displays the ‘Tea Rhythm East’ logo; the three ‘main display images’ in the middle showcase high-definition photography of jasmine tea, oolong milk tea, and osmanthus wine; the ‘artistic promotional slogan’ section features calligraphy writing ‘Eastern herbs, pure nature’; the bottom left ‘content area’ contains text introducing the ingredients, while the bottom right ‘user review section’ displays five-star ratings and short reviews like ‘Tea aroma is rich, sweet aftertaste!’.
You can also create advertising images based on sketches:
Prompt:Create an advertisement based on this sketch.
It can even draw on a specified grid.
Prompt:[Upload grid image] Generate a pixel art sprite of a traditional Chinese dragon head, perfectly aligned and filling this grid image. Please use only high-saturation red, yellow, and green color palettes, avoiding intermediate tones to ensure clarity.
💡 Exclusive Tips: How to Write Prompts?
After seeing so many examples, you may have noticed: NBP does not need those messy “spell” tags at all.The Google official documentation repeatedly emphasizes one “golden rule”: Speak to AI as if you are talking to a person.❌ Incorrect example (tag piling): “Cool car, neon, city, night, 8k, masterpiece, best quality.” (This is coding, not communication.)✅ Correct example (natural language): “A cinematic wide-angle shot of a futuristic sports car speeding through the rainy streets of Shanghai at night. The reflections of neon lights are mirrored on the wet pavement and the car’s metallic chassis.”Three suggestions for everyone:
- Don’t be too vague: “A woman” is an ineffective instruction; “A young Asian woman with a determined gaze” is an effective instruction.
- Learn to “fine-tune”: Don’t redraw every time. If an image is 80% satisfactory, tell it: “Great, change the lighting to sunset, leave everything else as is.”
- Provide background information: It has a brain, so tell it “You are designing for a bookstore”, and it will logically fill in many details.
In conclusionThe deeper I explore, the more I feel that NBP is unfathomable.To be honest, even though I have thoroughly reviewed the official documentation over the past two days, as I write this, I still feel that my understanding of it might just be passing.NBP is like a vast underground treasure trove; today’s article is at most a simple “treasure map” I brought back. The real treasure is still buried deeper, waiting for you in front of the screen to dig out the first shovel.To ensure you don’t leave empty-handed, I have bundled 20 sets of high-definition original images + complete prompts in the backend.👉 Reply with the keyword 【nbp1130】 in the backend to directly send you the link to the cloud drive.
If you also want to join this “treasure hunt”, remember to like, ♥️ let me know, and share it with your friends, and we will continue to dig deeper in the next issue![Reference]
https://x.com/GoogleAIStudio/status/1994480371061469306
The most comprehensive test of Nano Banana Pro: Strong to the point of absurdity!
The knowledge explanation diagram of Nano Banana Pro is definitely the most powerful update in text-to-image generation!
With Nano Banana Pro being so popular, how can it be used in China?
These amazing ways to use Nano Banana Pro, you definitely don’t know.