0成本打造亚马逊套图,教程
想系统掌握亚马逊广告的投放逻辑与底层闭环?
首先,你得有一个谷歌账号,没有的话手机注册一个就好。
然后进入谷歌的ai实验室:
https://aistudio.google.com/prompts/new_chat
里面包含了谷歌最前沿的各种ai工具:
对话,文生图,文生视频,文生音频。大部分都有免费使用额度。

文生图,目前有3种模型:Nano Banana, Imagen 4, Imagen 4 Ultra.
只有Nano Banana目前是免费的,默认选择它就好。
如果你进入工作台,发现默认模型不是Nano Banana。只需要点击右上角的红框区域。选中Nano Banana即可。


接下来便开始我们今天的正文,如何利用Nano Banana生成亚马逊套图。
今天的参考案例是一款Best Seller的宠物剃毛器。
emmm,其实我有一款差不多的剃头推子。。。还比狗用的便宜。。。
下面的图片是AI生成,和大卖前台的对比图,
大家可以先盲猜一下,哪个是Nano Banana作的。

全程用时3个小时。主要是用在指令构思和完善上。
实际生图只用了10分钟不到。
指令我更建议大家自己写,因为你在写的过程中,会有很多新的灵感,不断优化自己之前的指令。
揭晓答案:
上排为Nano Banana生成。
不知道跟大家想的是否一样。
由于是教学演示,生成这样一套图,只用到了一张产品主体并不完整的主图(只有正面)。
实际运用中,我们应该上传多视角图,方便AI抓取每个细节特征。
以保证生图结果更贴合实际产品。

具体如何生成呢?我们继续。
1.收集产品评论(非必须)
大家随便用什么工具都行,建议最终用txt文本格式,不要包含一些乱七八糟的跟产品无关的信息。方便AI分析。

2.让AI生成作图指令(仅供参考)
Generate 5 Amazon image generation prompts for the Nano Banana model, based on the requirements and materials provided below.
1. Data Source and Analysis Requirements:
Analyze File Content: Analyze the uploaded file to extract key customer pain points and selling points.
Combine with Five Points: Integrate the file analysis results with the "Five Points" below to ensure comprehensive selling point extraction.
2. Final Output Requirements:
Quantity and Type: Generate 5 complete image generation prompts, specifically: 2 Outdoor Lifestyle Scenes, 2 Feature Images, and 1 Indoor Lifestyle Scene.
Language and Aspect Ratio: All prompts must be in English, and the image aspect ratio must be fixed at 1:1.
Format: All prompts must strictly adhere to the "Structured Prompt Template" below, ensuring all information (especially the text embeddings) is clear and complete.
3. Selling Points and Parameter Text Embeddings (Must use the exact English text and precise layout below, ensuring only ONE text overlay per image):
All text embedded in the images must utilize the Icon + Bold Title + Detailed Description style, arranged either vertically stacked or corner-aligned as a single, distinct text layer
4. Core Product Information (Five Points):
Below 50 dB: The Oneisall silent motor reduces noise by 17%, ensuring that even the most sensitive pets won't be scared, allowing them to enjoy a comfortable grooming experience
Designed for Beginners: It comes with 6 guide combs(3/6/9/12/15/18mm) and a charger that allows for use while charging. With a one-touch start-up in just 1 second. Quick to master, show off your pet's fresh new look to the neighbors—crafted by you!
Sharp Yet Safe: The upgraded stainless steel fixed blade is sharper. The tool features a three-layer protection design to ensure pet safety during use, minimizing the risk of injury
Easy Maintenance: The cutter head can be removed in one second, and cleaning is just a matter of rinsing. This keeps the unit in pristine condition for consistent, optimal performance, ready for the next use
Continuously Top-Selling for 10 Years: Keeping the beautiful fur of pets in over 1.5 million households, our product is the number one choice in the market, trusted and loved by a wide range of pet enthusiasts
Note: For pets with long and thick hair, a preliminary trim with the included scissors can enhance the grooming experience!
5. Structured Prompt Template (Must be integrated):
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.
给大家拆解一下我的指令思路:
让AI根据我上传的评论内容去分析买家真实诉求,并结合前台实际五点,按照结构化指令模板,去生成5张亚马逊套图。
我并没有给Ai参考图,让它去图生图这种。因为这样没特色,会限制AI的发挥空间。
分析评论内容是为了更好的了解买家真实诉求,方便AI制作出更能击中买家诉求的场景图。
把五点融入进去,是为了我们有修图需要的时候,更好的结合我们优化后埋入的流量词。
结构化指令模板,则是参考的Nano Banana官方文档给出的最佳示例。
3.Nano Banana作图
等AI生成作图指令以后,就可以给Nano Banana逐一发送生图指令。
比例记得选1:1。或者4:3, 3:4。总之不要AUTO,保持图片尺寸一致性比较好。

注意,逐一发送很重要。因为:
1.每次最多只能生成4张,
2.逐一发送,能够针对某张图单独做细节微调,而不会影响其他已经生成的图

虽然也可以把所有指令一次性丢给它,让它直接生成套图。但这种傻瓜式的操作,意味着对多样性,精准性的妥协。还是不要盲目迷信一键生成套图。除非你是铺货,对图片没太大的要求。
另外,大家肯定也很关心细节到底如何,以下是实机图:




整体来说,我个人觉得还是可以,差不多堪用了。
包括它的排版,构图,设计。对场景化的理解和情感传达。
当然,细看还是有些微瑕疵:
比如第2张图的“CHARGE”挤到了一起。
第4张图,不同规格推子头上的小字挤到了一起。
但稍微PS加工一下,还是很好解决。(你没看错,有些细节处理上,Nano Banana暂时还是无法胜任,得PS加工)


首先,你得有一个谷歌账号,没有的话手机注册一个就好。
然后进入谷歌的ai实验室:
https://aistudio.google.com/prompts/new_chat
里面包含了谷歌最前沿的各种ai工具:
对话,文生图,文生视频,文生音频。大部分都有免费使用额度。

文生图,目前有3种模型:Nano Banana, Imagen 4, Imagen 4 Ultra.
只有Nano Banana目前是免费的,默认选择它就好。
如果你进入工作台,发现默认模型不是Nano Banana。只需要点击右上角的红框区域。选中Nano Banana即可。


接下来便开始我们今天的正文,如何利用Nano Banana生成亚马逊套图。
今天的参考案例是一款Best Seller的宠物剃毛器。
emmm,其实我有一款差不多的剃头推子。。。还比狗用的便宜。。。
下面的图片是AI生成,和大卖前台的对比图,
大家可以先盲猜一下,哪个是Nano Banana作的。

全程用时3个小时。主要是用在指令构思和完善上。
实际生图只用了10分钟不到。
指令我更建议大家自己写,因为你在写的过程中,会有很多新的灵感,不断优化自己之前的指令。
揭晓答案:
上排为Nano Banana生成。
不知道跟大家想的是否一样。
由于是教学演示,生成这样一套图,只用到了一张产品主体并不完整的主图(只有正面)。
实际运用中,我们应该上传多视角图,方便AI抓取每个细节特征。
以保证生图结果更贴合实际产品。

具体如何生成呢?我们继续。
1.收集产品评论(非必须)
大家随便用什么工具都行,建议最终用txt文本格式,不要包含一些乱七八糟的跟产品无关的信息。方便AI分析。

2.让AI生成作图指令(仅供参考)
Generate 5 Amazon image generation prompts for the Nano Banana model, based on the requirements and materials provided below.
1. Data Source and Analysis Requirements:
Analyze File Content: Analyze the uploaded file to extract key customer pain points and selling points.
Combine with Five Points: Integrate the file analysis results with the "Five Points" below to ensure comprehensive selling point extraction.
2. Final Output Requirements:
Quantity and Type: Generate 5 complete image generation prompts, specifically: 2 Outdoor Lifestyle Scenes, 2 Feature Images, and 1 Indoor Lifestyle Scene.
Language and Aspect Ratio: All prompts must be in English, and the image aspect ratio must be fixed at 1:1.
Format: All prompts must strictly adhere to the "Structured Prompt Template" below, ensuring all information (especially the text embeddings) is clear and complete.
3. Selling Points and Parameter Text Embeddings (Must use the exact English text and precise layout below, ensuring only ONE text overlay per image):
All text embedded in the images must utilize the Icon + Bold Title + Detailed Description style, arranged either vertically stacked or corner-aligned as a single, distinct text layer
4. Core Product Information (Five Points):
Below 50 dB: The Oneisall silent motor reduces noise by 17%, ensuring that even the most sensitive pets won't be scared, allowing them to enjoy a comfortable grooming experience
Designed for Beginners: It comes with 6 guide combs(3/6/9/12/15/18mm) and a charger that allows for use while charging. With a one-touch start-up in just 1 second. Quick to master, show off your pet's fresh new look to the neighbors—crafted by you!
Sharp Yet Safe: The upgraded stainless steel fixed blade is sharper. The tool features a three-layer protection design to ensure pet safety during use, minimizing the risk of injury
Easy Maintenance: The cutter head can be removed in one second, and cleaning is just a matter of rinsing. This keeps the unit in pristine condition for consistent, optimal performance, ready for the next use
Continuously Top-Selling for 10 Years: Keeping the beautiful fur of pets in over 1.5 million households, our product is the number one choice in the market, trusted and loved by a wide range of pet enthusiasts
Note: For pets with long and thick hair, a preliminary trim with the included scissors can enhance the grooming experience!
5. Structured Prompt Template (Must be integrated):
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.
给大家拆解一下我的指令思路:
让AI根据我上传的评论内容去分析买家真实诉求,并结合前台实际五点,按照结构化指令模板,去生成5张亚马逊套图。
我并没有给Ai参考图,让它去图生图这种。因为这样没特色,会限制AI的发挥空间。
分析评论内容是为了更好的了解买家真实诉求,方便AI制作出更能击中买家诉求的场景图。
把五点融入进去,是为了我们有修图需要的时候,更好的结合我们优化后埋入的流量词。
结构化指令模板,则是参考的Nano Banana官方文档给出的最佳示例。
3.Nano Banana作图
等AI生成作图指令以后,就可以给Nano Banana逐一发送生图指令。
比例记得选1:1。或者4:3, 3:4。总之不要AUTO,保持图片尺寸一致性比较好。

注意,逐一发送很重要。因为:
1.每次最多只能生成4张,
2.逐一发送,能够针对某张图单独做细节微调,而不会影响其他已经生成的图

虽然也可以把所有指令一次性丢给它,让它直接生成套图。但这种傻瓜式的操作,意味着对多样性,精准性的妥协。还是不要盲目迷信一键生成套图。除非你是铺货,对图片没太大的要求。
另外,大家肯定也很关心细节到底如何,以下是实机图:




整体来说,我个人觉得还是可以,差不多堪用了。
包括它的排版,构图,设计。对场景化的理解和情感传达。
当然,细看还是有些微瑕疵:
比如第2张图的“CHARGE”挤到了一起。
第4张图,不同规格推子头上的小字挤到了一起。
但稍微PS加工一下,还是很好解决。(你没看错,有些细节处理上,Nano Banana暂时还是无法胜任,得PS加工)







福建
12-12 周五











