你还在为 App Store 截图花上几个小时而头疼吗?
我太懂这种挣扎了。
我当时拼命想找个办法,又不想花钱买那些 AI 截图工具,于是在 ChatGPT 图像生成更新后试了试它,说真的?结果把我惊艳到了。
这是我总结的 5 步流程,从此再也不用为截图发愁。
第一步:给 ChatGPT 打好基础
如果你是用 AI agent 开发的应用,让它生成一份快速的 TLDR 交接文档,突出你应用的核心功能与最佳营销角度。如果是你自己开发的,就自己写一小段话,不用太复杂。
然后把这段话粘贴到 ChatGPT 里,让它读一下并告诉你它理解了什么。这一步很重要:你要确保它在你开始做创意决策之前,真正*理解*你的应用。
第二步:制定截图方案
现在让 ChatGPT 制定 两个版本 的 8–10 张截图方案。告诉它重点关注:
- 营销导向的文案(好处,而非功能)
- 简洁有力的标题,便于快速浏览
- 截图流程以你最重要的 3 个卖点开头
- 建议在每张幻灯片中展示哪些具体的 UI 界面
一个很有效的框架技巧:开头写 *“你是一位资深应用开发者,拥有敏锐的营销和视觉设计眼光……”* —— 这会明显改变输出的语气。
你可以直接套用下面这个框架,效果最好:
You are an App Store conversion expert. I'm going to give you my app's description, and I need you to create a screenshot flow for App Store Connect.
Your job:
Analyze the description and generate a plan for 8–10 App Store screenshots, ordered strategically for maximum conversion. The first 3 screenshots are the most critical, they appear in search results before the user taps in, so they must immediately communicate the core value and create curiosity or urgency.
For each screenshot, give me:
- Screenshot #: (1–10)
- Marketing angle: What emotional or practical hook this slide targets
- Headline: Short, punchy (5 words or fewer ideally)
- Subtext: 1–2 supporting lines that add context or social proof
- What's shown on screen: Brief description of the UI or visual being displayed
Rules:
- Screenshots 1–3 must hook, differentiate, and create desire, no fluff
- Progress through the flow logically (hook → features → proof → CTA)
- Use the app's actual language and terminology from the description
- Prioritize benefits over features
- The final screenshot should drive action (download, start free, etc.)
Here is my app description:
[PASTE YOUR APP DESCRIPTION HERE]
Output the full 8–10 screenshot plan in the format above. Do not skip any fields.如果第一版输出你不满意,就一起调整。修改文案、调整流程顺序、对任何感觉不对的地方提出意见。目标是协作,而不是甩手不管。
第三步:生成截图
方案定稿后,用下面这个移动端前端设计 skill 生成每张图片。这个 skill 才是重头戏。
这是该 skill 的直接链接,你可以复制粘贴,或者下载后以文件形式发送给 ChatGPT。我推荐后者,因为内容比较长。
https://github.com/Leonxlnx/taste-skill/blob/main/skills/imagegen-frontend-mobile/SKILL.md
这个 skill 对我来说简直是神器,原分享者 @blueemi99 功劳全归他。除了截图,它还能很好地用于应用原型、引导流程和其他前端视觉设计。这个以后再讲。
接下来告诉 ChatGPT 你想从哪张幻灯片开始、用哪些文案,以及任何视觉偏好:
- 色板
- 风格(简约、大胆、渐变等)
- 角色或吉祥物(如果你的应用有的话)
如果你有品牌素材、图标、吉祥物、灵感图片,也一并附上。给的上下文越多,输出效果越好。
下面是我会用的精确提示词:
Great, I love this plan. Let's generate Screenshot [NUMBER] now.
Please create the App Store screenshot image at 1242 × 2688 px (the required App Store Connect upload size). Portrait orientation, PNG format.
For this image:
- Display a realistic iPhone mockup with my actual app UI shown on screen (I'm attaching a screenshot of the real app, use this as the screen content inside the mockup)
- Place the headline "[HEADLINE FROM PLAN]" prominently, large and legible
- Place the subtext "[SUBTEXT FROM PLAN]" below it in a smaller weight
- Background: [describe a color, gradient, or say "match the app's color palette"]
- Style: clean, modern, minimal, similar to top wellness/productivity apps on the App Store
Do not add badges, watermarks, or decorations not mentioned above. Keep it conversion-focused and uncluttered.
[ATTACH YOUR REAL APP SCREENSHOT HERE]第四步:调整尺寸以符合 App Store Connect 要求
ChatGPT 生成你满意的图片后,它出图时几乎肯定不会是 App Store Connect 所需的正确尺寸。这很正常,别慌。
用下面这个跟进提示词来获得可下载的文件:
The image you generated doesn't upload correctly to App Store Connect — it's rejecting it due to incorrect dimensions. The file I downloaded is not actually 1242 × 2688 px even though that's what I asked for.
Please do the following:
1. Take the screenshot design you just created
2. Resize it to exactly 1242 × 2688 pixels — do not crop, just resize the canvas to these exact dimensions
3. Export it as a PNG file
4. Provide it as a downloadable link or file attachment so I can download it directly and upload to App Store Connect
The exact required spec is:
- Width: 1242 px
- Height: 2688 px
- Format: PNG
- Color space: RGB
- No alpha channel请先确认文件的像素尺寸,再分享出去。
让 ChatGPT 以 **1242 × 2688 px** 导出图片,并提供可下载的 PNG 文件。然后直接在 App Store Connect 中上传测试。如果能通过,大功告成,进入下一步。如果被驳回,把错误信息粘贴回 ChatGPT,告诉它文件不行。通常试一两次就能搞定。

在 App Store Connect 中测试文件上传是否成功
## **第 5 步:为每张截图重复操作**
以上就是完整流程。回到第 3 步,处理下一张幻灯片,重复操作,直到所有截图完成。
全程跟 ChatGPT 协作,告诉它你喜欢什么、不喜欢什么、需要改什么。你反馈得越多,输出的质量就越高。

完整的 ChatGPT 截图生成流程
就说这么多。希望能帮你省下几小时(以及一堆抓狂的时刻)。
创作愉快 🙂