Recently, the AI image generation model “Nano Banana” (Gemini 2.5 Flash Image) has gained immense popularity worldwide. Google quickly released an official prompt guide, sparking a wave of practical applications on platforms like Xiaomoshub, particularly in generating realistic figurine images that have taken the global market by storm.
1. Background and Core Features of Nano Banana
Nano Banana initially appeared anonymously on the LMArena platform and became popular due to its outstanding image generation capabilities. Google later claimed it as the Gemini 2.5 Flash Image model. Its five core features include:
-
Text-to-image: Generate high-quality images from text descriptions.
-
Image + text-to-image (editing): Add, remove, or modify elements through text prompts, changing styles and colors.
-
Multi-image to image (composition and style transfer): Create new scenes or transform styles using multiple input images.
-
Iterative refinement: Fine-tune images through multiple dialogues until perfection is achieved.
-
High-fidelity text rendering: Generate clear and well-structured text images suitable for logos, charts, and posters.
Google specifically advises: When describing a scene, do not just list keywords. Compared to simple keywords, narrative-rich and detailed paragraphs can generate more coherent images.
2. Official Prompt Guide: 6 Categories of Templates and Effects
Google summarizes six categories of prompts to help users easily navigate Nano Banana. Each category provides prompt templates, code examples, and effect images.
2.1 Realistic Scenes
To obtain realistic images, think like a photographer by specifying camera angles, lens types, lighting, and details.
Prompt template:
A photorealistic [shot type] of [subject], [action or expression], set in [environment]. The scene is illuminated by [lighting description], creating a [mood] atmosphere. Captured with a [camera/lens details], emphasizing [key textures and details]. The image should be in a [aspect ratio] format.
Code example:
Effect image: Soft golden sunlight illuminating the details of clay and the wrinkles of an elderly person.
2.2 Stylized Illustrations and Stickers
Suitable for creating stickers, icons, and other materials, requiring specification of style and background.
Prompt template:
A [style] sticker of a [subject], featuring [key characteristics] and a [color palette]. The design should have [line style] and [shading style]. The background must be white.
Code example: Same structure as the previous section, replacing prompt content.
Effect image: Kawaii style bear sticker with a clear white background.
2.3 Accurate Text in Images
When rendering text, describe the content, font, and design.
Prompt template:
Create a [image type] for [brand/concept] with the text “[text to render]” in a [font style]. The design should be [style description], with a [color scheme].
Code example:
Effect image: Black and white minimalist logo with a prominent coffee bean icon.
2.4 Product Models and Commercial Photography
Create professional product images for e-commerce or advertising.
Prompt template:
A high-resolution, studio-lit product photograph of a [product description] on a [background surface/description]. The lighting is a [lighting setup, e.g., three-point softbox setup] to [lighting purpose]. The camera angle is a [angle type] to showcase [specific feature]. Ultra-realistic, with sharp focus on [key detail]. [Aspect ratio].
Code example:
Effect image: Product photo of a coffee cup with realistic steam details.
2.5 Minimalism and Spatial Design
Suitable for website or poster backgrounds.
Prompt template:
A minimalist composition featuring a single [subject] positioned in the [bottom-right/top-left/etc.] of the frame. The background is a vast, empty [color] canvas, creating significant negative space. Soft, subtle lighting. [Aspect ratio].
Code example:
Effect image: A maple leaf in the bottom right corner, leaving blank space suitable for text overlay.
2.6 Comic Strips (Comic Panels/Storyboards)
Create storyboards or comics.
Prompt template:
A single comic book panel in a [art style] style. In the foreground, [character description and action]. In the background, [setting details]. The panel has a [dialogue/caption box] with the text “[Text]”. The lighting creates a [mood] mood. [Aspect ratio].
Code example:
Effect image: A black-style comic with a detective atmosphere under a streetlight.
3. Practical Applications: Figurine Generation and Creative Uses
Nano Banana has sparked a figurine image craze on Xiaomoshub, with users amazed by the practical results. Here are key applications and prompt templates.
3.1 Figurine Generation Practical Test
Prompt template:
Use the nano-banana model to create a 1/7 scale model, 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 TAMIYA-style toy packaging box printed with the original artwork.
(Translation: Use the nano-banana model to create a 1/7 scale physical model, maintaining a realistic style and environment. Place the model on a computer desk, with a circular transparent acrylic base and no text. The computer screen shows the modeling process of the model in ZBrush. Next to the computer screen, place a TAMIYA-style toy packaging box printed with the original illustration.)
Practical test results:
-
Anime characters like Eren Yeager, with accurate model recognition and rich details.
-
Non-full-body images (like Gintama characters) may generate abnormal parts, but overall reasonable.
-
Pet generation: Cat and dog figurines are realistic and cute, with users asking for purchase links.
-
Full-body images of real people: Quirky poses can also be perfectly replicated, suitable for generating funny figurines of friends.
Improvement tip: If the model appears to be “floating”, add support descriptions, such as “add a transparent stand to the model”.
3.2 Other Popular Applications
Users explore various creative applications, keeping the prompt templates intact:
-
Pose control: Combine multiple images to generate specified poses.
Prompt: Have the characters in image one and image two pose like in image three, with one arm resting on the other’s shoulder and the other arm making a heart gesture.
Effect image: Character interaction poses.
Prompt: Have the characters in image one and image two fight in the pose from image three, with one throwing a punch and the other kicking.
Effect image: Fighting dynamics.
-
Animation generation: Create coherent animations by combining video models (like Seedance) with sketches.
-
Breaking the dimensional wall: Generate real-life cosplay.
Prompt: Generate a photo of a real person cosplaying this illustration, with the background set to Comiket.
Effect image: Cosplayer image with a comic convention background.
4. Technical Details and Future Directions
The team revealed the key technologies behind the success of Nano Banana:
-
Text rendering as a core metric: Replacing subjective human preference assessments with objective, quantifiable standards. When the model handles complex text structures well, it can also learn other image structures.
-
Native multimodal and interleaved generation: Deeply integrating image understanding and generation capabilities to achieve “forward transfer”. Interleaved generation supports multi-round editing, breaking complex tasks into steps (e.g., completing 50 editing requests step by step).
-
User feedback-driven improvements: Collecting failure cases from social platforms (e.g., pixel-level editing inaccuracies, poor character consistency) to build internal testing benchmarks. The new model addresses legacy issues from Imagen 2.0, such as more natural image textures and multi-angle character rendering.
Future goal: Integrate all modalities into Gemini to achieve AGI.
5. Events and Experience Opportunities
Google announced the Nano Banana Hackathon, lasting two days, offering free API credits. Winners can receive prizes, developer points, and Gemini-related products.
Experience opportunities:
-
AI Studio: Directly use the Gemini 2.5 Flash Image model.
-
Gemini API: Integrate into custom applications.
AI Studio link: https://aistudio.google.com/prompts/new_chat?model=gemini-2.5-flash-image-preview
Reference links:
-
Google official guide: https://ai.google.dev/gemini-api/docs/image-generation
One-click three connections: “Like”, “Share”, “Heart”, and feel free to share your experiences in the comments!
Custom figurines from photos: Photo-customized figurines
Custom figurines from photos: Photo-customized figurines