跳到正文
DEV Community· Катерина·· 3 小時前AI 評分53

我如何調整 AI 生成廣告尺寸而不破壞原有設計

How I Resize AI-Generated Ads Without Ruining the Original Design

AI 導讀

作者分享如何透過提示詞調整 AI 生成廣告的尺寸,同時保留已核准設計中的產品、包裝、文字和品牌視覺。她以桌面版 2880 × 1024 px 改成手機版 750 × 1108 px 為例,建議明確列出必須保留及可以調整的元素,並要求 AI 無法自然適配時説明限制。生成後仍須核對產品包裝、文字、構圖、品牌一致性及匯出尺寸,必要時可在 Canva 手動修正。

正文

I've been working with advertising creatives for almost a year now, and during that time, I've learned a few things about using AI for commercial design.

One of the biggest challenges I faced was adapting the same advertising creative to different sizes without changing the original design.

You create a banner, everything looks good, the client approves it, and now you need a mobile version.

Sounds simple, right?

But AI doesn't always understand that you want to change the layout, not the design itself. Products can look different, packaging details can change, and sometimes the entire composition gets redesigned.

I've run into this more than once.

After experimenting with different prompts, I found an approach that makes the whole process much more predictable.

And it comes down to one simple thing: telling AI exactly what it can change and what it needs to leave alone.

Here's how I do it

Let's say I've created a desktop banner for a promotional campaign, and now I need to adapt it for mobile.

For example:

  • Desktop banner: 2880 × 1024 px
  • Mobile banner: 750 × 1108 px

These are two completely different layouts.

Simply stretching or cropping the original image won't work. The products need to be repositioned, the text needs to remain readable, and the entire composition has to make sense in a vertical format.

But here's the important part: I don't want AI to create a new design.

I want it to work with the one I already have.

So instead of writing something like "Resize this banner for mobile," I upload the original image and use a prompt that clearly separates what must stay the same from what can be adjusted.

The prompt I use

ROLE

Act as a senior advertising designer specializing in e-commerce visuals and professional layout adaptation.

CONTEXT

I'm uploading an existing advertising creative that needs to be adapted to a different format.

The original design is already approved. Your job is to preserve its visual identity while adapting the composition to the new dimensions.

TASK

Adapt the uploaded image to:

TARGET DIMENSIONS: [INSERT WIDTH × HEIGHT]

Use the original image as the main visual reference.

CONSTRAINTS

PRESERVE:

  • Exact products, packaging, labels, and logos
  • All original promotional text and spelling
  • Brand colors, lighting, and visual style
  • Product proportions and important details
  • The overall visual identity of the campaign

YOU MAY CHANGE:

  • Canvas dimensions and aspect ratio
  • Positioning and spacing of existing elements
  • Text placement to improve readability
  • Background extension when necessary
  • Composition to fit the new format naturally

DO NOT:

  • Redesign the original creative
  • Invent, replace, or remove products
  • Change packaging or product details
  • Add new decorative elements
  • Rewrite promotional text
  • Stretch or distort the original image

OUTPUT FORMAT

Create one adapted version of the original creative for the requested format.

Keep the composition visually balanced and professional.

If an element cannot fit naturally without changing an important detail, explain the limitation instead of silently modifying it.

One more thing I've learned

A detailed prompt helps, but it doesn't guarantee a perfect result.

AI can still make small mistakes, especially with product packaging, labels, and typography.

That's why I always compare the adapted version with the original before sending anything to a client.

I usually check four things:

  1. Products: Are the shapes, colors, and packaging details still correct?
  2. Text: Is every word accurate and readable?
  3. Composition: Does the new layout actually work for its intended format?
  4. Brand consistency: Does it still look like part of the same campaign?

If something needs a small correction, I often find it easier to fix it manually in Canva rather than regenerate the entire image.

I also check the final export dimensions, because requesting a specific size from an AI image tool doesn't always mean the generated file will have those exact pixel dimensions.

My takeaway

I've learned that getting better results with AI isn't always about finding a better tool.

Sometimes it's simply about being more specific with your instructions.

For me, separating the things AI can change from the things it absolutely needs to preserve has made adapting advertising creatives much easier to manage.

It doesn't mean every image comes out perfect on the first try. But it gives me more control over the process and helps me avoid unnecessary redesigns.

And that's exactly what I need when working on real client projects.

If you're also working with AI tools, design, content, or digital projects, I share more practical prompts and workflows in my Telegram channel, WORKKIT.

https://t.me/workkitai

Just useful AI tricks and tools that make everyday work a little easier.

來源:DEV Community · dev.to