如果你暂时请不起美工，也不懂拍摄，甚至PS水平也不行。那么这篇文章应该对你有帮助。但你也知道，电商很大程度上是在做视觉表达。更直白一点，就是在卖图片。很多朋友的listing点击率早就已经达标了，差就差在图片上。虽然之前的文章，已经演示过如何利用nano banana生成产品图。但由于很多朋友没能掌握提示词表达，不能很好的生成多场景提示词。比如这样一张场景图，按以前的做法是先抠图，再排版设计。目前市面上倒是有一些积分制的AI图片翻译工具，但小卖家其实也用不起。如果你有PS功底，想自己排版设计，那么可以用下面的提示词:Strictly remove only overlay text elements from the image. Do not remove, distort, or smooth any non-textual elements, including product details, textures, lines, patterns, logos (if not text), or background features. The final result must be a clean, text-free image that perfectly retains all original product fidelity, color, and sharp resolution.尽情发挥PS创意去吧。
但据我所知，大部分朋友对自己的PS功底也是没信心的。
那就更简单粗暴点，让Nano banana直接做裂变。提示词如下：
First, analyze the input image to identify the main product object and extract all key selling points. STRICTLY preserve the main physical product object(s) exactly as they are, maintaining its original shape, viewing angle, and proportions; do not alter the product itself. Completely discard the original background, layout, text, and graphics. Replace the environment with a brand new, premium, modern context optimized for high-conversion Amazon listings, ensuring the new scene seamlessly integrates with the product's unchanged appearance and perspective.
Conditional Logic for Models: STRICTLY ONLY if a human model exists in the original image, transform their appearance to a professional Western/Caucasian model integrated into the new scene. IF NO HUMAN exists in the original, DO NOT ADD A PERSON; keep the focus solely on the product.
Content Strategy: Translate all concepts into professional, persuasive English marketing copy. CRITICALLY CONSOLIDATE information to eliminate ALL redundancy. Each distinct data point or benefit must be presented ONLY ONCE in the final design (e.g., do not repeat the same data point in both a headline and an icon bubble). Redesign infographics using sleek, minimalist vector icons and a fresh typography hierarchy arranged in a clean, dynamic layout distinct from the original. Do not add other brands.
这样一来，你就会得到类似下面的图：
基本原则是：[保留什么] + [移除什么] + [不要什么]的Do & Do not结构.1.如果对当前图片不满意，不需要一直输入提示词。直接点击提示词或者图片右上角的星星标志，就可以重新生图油猴插件使用教程，请务必自行百度。否则真的会拉黑你。a.生成图片后，先用最开始的提示词，移除所有文字+图标内容;
3557
