In the past couple of days, the group has been researching how to use Nano Banana to generate 3:4 cover images for Xiaohongshu with one click. Today, we finally developed it! I have gathered insights from Master Cang, A Zhen's prompts, and Super Huang's requirements to create a solution that can consistently output Xiaohongshu cover cards. After testing it 10 times, it has finally stabilized. You can take a look at the image effects. Highlights of this solution: Separation of text and images, each serving its purpose: Since Nano Banana's Chinese is not great, we let Nano Banana focus on generating high-quality illustrations, and then use HTML code to accurately "stick" the text on. The creativity of AI and the stability of code do not interfere with each other. Moreover, this solution not only supports Nano Banana but also all drawing models; the three images below in the six-grid layout are from Flux kontext. Goodbye to the randomness of text placement: This completely solves the pain point of AI writing with random positions and ugly styles. The title position, font, and size are all hard-coded, ensuring that each image is professional and uniform. Template-based, one-click reuse: Once the layout is finalized, it becomes a universal template. In the future, if you change the topic, you only need to modify the requirements, and the style will remain consistent, making it very suitable for building a personal brand tone.