The Popular Image Generation Tool: How to Use ‘Nano-Banana’ Indefinitely?

Google「Nano–Banana」Image Generation / Editing Model Explained: What can it do, where can it be used, and how to create 3D figurines?

Nano-Banana has been trending for over half a month, but due to various reasons including health, I haven’t released related videos and articles. Today, I will cover everything at once. After reading, you will be able to theoretically use this very powerful image generation model indefinitely.

nano-banana

1. What is Nano-Banana?

Google officially launched Gemini 2.5 Flash Image in its update on August 2025, referring to the image generation and editing model as “nano-banana”: it supports text-to-image generation, image + text editing, multi-image blending/style transfer, and conversational local modifications.

• Release and capability overview: Developers Blog; Gemini Product Blog (emphasizing character/pet consistency, clothing changes, blending, style application, and other scenarios).

Community Popularity: Google executives disclosed that after integrating Nano-Banana, the Gemini App gained over 10 million new users and over 200 million edits; “turning characters/pets into 3D figurines” has become a viral trend.

2. What are its strengths (core advantages)

• Continuous editing & character consistency: Even after multiple modifications, it can maintain facial/features consistency, avoiding the issue of “the more you edit, the less it resembles the original”.
• Conversational control: Directly use natural language to “add/remove objects, change materials/lights/poses”, and also leverage the model’s world knowledge to assist in generation and editing.
• Multi-image to image (Blend): Blend multiple input images into a unified scene, or transfer the style of image A to image B.
• Text rendering readability: Chinese/English titles and fonts are more stable, suitable for posters and UI screens.
• Traceability and compliance: By default, it embeds a SynthID digital watermark (invisible), which helps platforms label “AI generated”.

The key point here is the first one—character consistency. If it were other capabilities, we are familiar with GPT-4o which can easily achieve them, even faster, but in terms of consistency, so far Nano-Banana remains unique.

3. Where can it be used (official/mainstream access | recommended to bookmark)

• Gemini Web Version (direct experience): https://gemini.google.com (select 2.5 Flash, you can upload reference images or write prompts directly)

• Google AI Studio (Playground): https://aistudio.google.com (select Gemini 2.5 Flash Image / Nano-Banana in the upper right corner; supports “one-click generate API code”)

• Google AI Lab Unlimited Use (Image generation with Gemini / Nano-Banana): https://labs.google/fx/zh/tools/whisk

4. Viral Gameplay: Turning People/Pets into 3D Figurines

Gameplay idea: Upload a clear half-body photo of a person/pet → set “1/7 scale figurine + transparent base + packaging box + desktop scene + shallow depth of field”; Nano-Banana has stable effects in facial consistency and material texture.

Chinese prompt (copy to use)

将我上传的照片转成 1/7 比例的「商业化实体手办」，写实风。摆在电脑桌上，使用无文字的透明圆形亚克力底座。请保证与原图的五官与发型一致；塑料材质要有哑光上色的笔触质感。旁边放一个玩具包装盒，盒面参考我的照片做海报风（不要出现品牌商标）。摄影级柔光，浅景深。

English prompt (community tone)

Create a 1/7-scale commercialized action figurine of the person in the reference photo. Place it on a computer desk with a clear round acrylic base (no text or logos). Keep facial identity consistent with the original photo. Realistic plastic with matte paint strokes. Add a toy box next to the monitor; the box art references the portrait without using trademarks. Shallow depth of field, studio soft lighting.

Extended style keywords (replace for different styles): <span>Funko Pop / Nendoroid / Q-style / clay / LEGO / wool felt / diorama / resin kit / garage kit</span>

5. More Highly Discussed Prompt Templates (copy to use)

1) Poster

Generate a 1080×1350 vertical poster, main title “Weekend Market” (currently, the availability of Simplified Chinese is very poor, you can try English prompts), leave space for subtitle and QR code; warm neon + soft light; clean white space.

This model has optimized text readability for English, suitable for designs with English titles.

2) Multi-image blending + style transfer

Blend two uploaded portraits into a group photo; maintain facial consistency of both individuals; apply an overall 1980s studio style with a velvet curtain background and film grain.

3) Image modification (local replacement)

Replace the T-shirt with a beige knitted cardigan while keeping the facial features unchanged; add backlighting and skin highlights; do not change the hairstyle.

6. Important Notes (must read)

• Copyright/Portrait/Trademark: For commercial materials, please ensure you have the corresponding authorization; “BANDAI style box” can serve as a style reference, but do not directly use others’ photos. Of course, the image of Sam I used above is just for demonstration purposes 🐶.
• Watermark labeling: Images generated/edited by Gemini will embed a SynthID; some interfaces will also add visible watermarks.
• Usability/Network: When the domestic network is unstable, it is recommended to use AI Studio first (or switch networks/DNS).

References and Further Reading

https://developers.googleblog.com/en/introducing-gemini-2-5-flash-image/

https://aistudio.google.com/models/gemini-2-5-flash-image

https://ai.google.dev/gemini-api/docs/image-generation

https://gemini.google/overview/image-generation/