Google’s newly released Nano Banana image model is going viral on the Internet right now.
After dozens of hours of hands-on use, I’ve put together a complete guide for you, covering: an introduction, core features, how to use it, and—most importantly—the most comprehensive tutorial on creative applications (with prompts included). Hope you enjoy it.
Let’s get started!
Nano Banana is Google’s newly released image generation and editing model, officially named Gemini 2.5 Flash Image.
The nickname “Nano Banana” originated from its anonymous performance on the LMArena blind testing platform, where its outputs occasionally included banana icons—leading users to affectionately refer to it as "Nano Banana."
Within just 10 days, the model generated over 200 million images, making it one of the most popular AI products of the year.
Its standout feature is the ability to maintain a character’s facial features, hairstyle, and clothing style across multiple edits. Even when the scene or pose changes, the character remains highly recognizable.
Up to three images can be merged at once, seamlessly combining materials into a natural-looking composition—ideal for product showcases or character design.
Supports precise edits using simple commands like “remove the wrinkles on the clothes” or “change the blue to green.”
Users can continuously fine-tune the output based on the previous result, gradually refining it into a final version they’re satisfied with.
Currently the most popular use case! Simply upload a photo to generate a realistic 3D figurine. Both real-life and anime-style images are supported. It can turn pictures into 3D figurines no matter anime or realistic photos.
Prompt examples:
“Use the nano-banana model to create a 1/7 scale commercialized figure of the character in the illustration, in a realistic style and environment. Place the figure on a computer desk, using a circular transparent acrylic base without any text.On the computer screen, display the ZBrush modeling process of the figure. Next to the computer screen, place a BANDAl-style toy packaging box printed with the original artwork.”
You’ll need to upload a Google Maps screenshot that includes a red directional arrow.
Prompt:
Prompt: You are a location-based AR experience generator. Highlight the [Point of Interest] in this image and annotate it with relevant information.
Note: Replace [POI] with the specific point of interest you want to label.
Input: A reference image containing the target object must be uploaded.
Prompt: Generate a daytime isometric view of the image [limited to buildings].
Note: You can modify the content inside the [brackets] as needed—it doesn’t have to be buildings; you can specify vehicles, characters, or other objects.
Input: A portrait photo of the person must be uploaded.
Prompt: Transform the character’s style into the classic [1990s] [male] look. Add [long curly hair], [a long beard], and change the background to the iconic [California summer scenery]. Adjust the photo style to resemble a [1990s Polaroid]. Do not alter the character’s facial features.
Note: Replace the [brackets] with your desired era and style details.
Input: Multiple reference images must be uploaded and merged into a single composite image.
Prompt: The model is posing against a Porsche, wearing the items shown in the reference images. The background is light gray, and a tiger is sitting next to the model.
Note: The prompt should include detailed descriptions and reference multiple visual elements. All reference images must be combined into one before uploading.
Input: Upload character images along with a hand-drawn sketch.
Prompt: Make these two characters fight using the pose shown right image. Add appropriate visual background and scene interaction. Generate the image in a 16:9 aspect ratio.
Input: A ground-level photo must be uploaded.
Prompt: Convert the photo to a top-down (bird’s-eye) perspective and mark the photographer’s location.
Input: Upload one sticker reference image and one character image.
Prompt: Turn the character into a white-outline sticker similar to Image 2. Convert the character into a web-style illustration, and add a playful white-outline phrase that describes Image 1.
Input: An illustration image must be uploaded.
Prompt: Generate a photo of a girl cosplaying the character in this illustration. Set the background at Comiket.
Input: A reference image of the character must be uploaded.
Prompt: Generate a full character design for me, including:
Costume Design (outfit breakdown and styling)
Input: Upload one line art image and one color palette image.
Prompt: Accurately apply the colors from the right to the character on the left
Input: Upload a portrait image of the person whose hairstyle you want to change.
Prompt: Generate a 3×3 grid showing this person with different hairstyles.
Prompt: Illustrate a [3D model showcasing the aircraft carrier] for academic presentation. Include detailed annotations and explanations to demonstrate its mechanism and the function of [each part]. The design should be highly realistic, anatomically accurate, and meticulously detailed.
Note: Replace the [brackets] with the specific model you want to showcase.
Input: Upload a photo containing a variety of ingredients.
Prompt: Use these ingredients to prepare a delicious lunch for me. Present the meal on a plate with a close-up view. Remove any other plates and leftover ingredients from the scene.
Input: An old, damaged photo must be uploaded.
Prompt: Restore and colorize this photo.
Input: Upload a character image and a clothing reference image.
Prompt: Replace the character’s outfit in the input image with the target clothing shown in the reference image. Maintain the character’s pose, facial expression, background, and overall realism. Ensure the new outfit looks natural, fits well, and matches the lighting and shadows. Do not alter the character’s identity or environment — only change the clothing.
Input: Upload a line art image and a reference image.
Prompt: Replace the character in Image 1 with the pose shown in Image 2. Render the result as if shot in a professional photo studio.
Prompt: Create an infographic showing the five tallest mountains in the world.
Input: A reference image must be uploaded.
Prompt: Photograph the product in a dramatic, modern setting with an explosive outward dynamic arrangement. Let the fresh, raw key ingredients fly around the product to highlight its freshness and nutritional value. This should be a promotional ad-style shot with no text, focusing entirely on the product. Use the brand’s key color as the background.
Input: A reference image must be uploaded.
Prompt: Based on the uploaded image, create a manga-style comic strip with added dialogue and an engaging storyline. I’d like it to be a Japanese manga.
Input: A reference map image must be uploaded.
Prompt: Use this location as a landmark and render it as an isometric image (buildings only). Apply a theme park game art style.
Input: A character reference image must be uploaded.
Prompt: Generate a four-panel illustration process for the character:
1. Line art
2. Flat colors
3. Add shading
4. Final refinement and rendering No text in the images.
Input: Upload one character reference image and one makeup reference image.
Prompt: Apply the makeup from Image 2 to the character in Image 1, while preserving the pose from Image 1.
Input: A character reference image must be uploaded.
Prompt: Create a pose sheet for this illustration, showcasing a wide variety of character poses.
Input: Upload one product reference image and one packaging reference image.
Prompt: Apply the product from Image 1 onto the can shown in Image 2. Place the result in a minimalist set design, styled as a professional photoshoot.
Input: A reference image must be uploaded.
Prompt: Transform the person in the photo into a LEGO-style minifigure and present it in the style of a LEGO packaging box. Use isometric perspective. Label the box with the title “Lego.” Inside the box, showcase a minifigure based on the person in the photo.
Prompt: Create an exploded diagram of an Apple iPhone, showcasing all its accessories and internal components.
Input: A food reference image must be uploaded.
Prompt: Annotate the meal with food names, calorie density, and approximate total calories.
Input: A reference image must be uploaded.
Prompt: Extract the model from the image and place her onto a transparent background.
Input: An ancient reference image must be uploaded.
Prompt: Convert the image into the style of a realistic aerial photograph taken today.
Input: A reference image must be uploaded.
Prompt: Create a fashion mood board collage. Surround the portrait with cut-out patterns of the clothing items worn by the model. Add handwritten notes and sketches in a playful marker-pen style, and label each item in English with its brand name and source. The overall aesthetic should be both creative and cute.
Prompt: A high-resolution advertising photograph of a man delicately holding a realistic miniature [product] between his thumb and forefinger. The background is clean and fresh, with studio lighting and soft shadows. The hand pose is refined, with natural skin tones, and positioned to highlight the product’s shape and details. The product appears extremely small yet richly detailed, with precise brand representation, centered in the frame, and a shallow depth of field. The style emulates luxury product photography and minimalist commercial aesthetics.
Replace the [brackets] with the product you want to showcase.
Input: A reference image must be uploaded.
Prompt: A realistic photographic scene. A giant statue of this person is placed in a central square in Tokyo, with people gazing up at it
Input: A character reference image and a scene composition reference image must be uploaded.
Prompt: Seat three characters on chairs in a restaurant.
Input: A reference image must be uploaded.
Prompt: Convert the input image into a black-and-white manga-style line drawing.
Input: A Google Maps reference image must be uploaded.
Prompt: Using this location, create an isometric image of the landmarks in the HD-2D style of the game Minecraft (buildings only).
Input: A reference image must be uploaded.
Prompt: Create merchandise using this character image.
As of now (September 2025), Google’s “Nano Banana” (officially known as Gemini 2.5 Flash Image) image generation and editing model is available through the following platforms:
A recommended option is using LMArena or Vidsoul, where Nano Banana is available for free. Gemini 2.5 Flash Image originally debuted anonymously on LMArena. This is a community-driven platform where models are placed into anonymous “Battle Mode,” allowing users to upload prompts or reference images and compare outputs side-by-side with another model, then vote on which is better.
You can experience Nano Banana directly on LMArena in Battle, Side-by-Side, or Direct Chat modes.
Google AI Studio: You can try Nano Banana for free (listed as “Gemini 2.5 Flash Image Preview” in the model selection menu). Simply choose it in AI Studio and use text prompts or image uploads to create and edit visuals.
Gemini API / Vertex AI: If you have an API key and have enabled billing, you can access the model via Gemini API or Vertex AI to build products or services. However, this requires API permissions or a Google Cloud setup. The billing structure is complex, and token usage needs to be carefully managed—those unfamiliar with it may easily run into issues.
Direct Use on Gemini: Free usage is available (with daily editing limits), and a premium subscription is offered at around $19.99/month. It’s convenient and straightforward: text prompts + image editing, with SynthID watermarking. Great for quick everyday experiments. However, free users face limits in usage frequency, output quality, and speed. Frequent use requires payment or is subject to daily caps.
As the world’s top-ranked image editing model, Nano Banana is transforming the creative industry landscape.
More than just a technical tool, Nano Banana serves as a catalyst for change in the creative sector, signaling that AI will achieve breakthroughs in an even wider range of professional fields, expanding from 2D imagery into 3D modeling, video generation, and beyond.
