This website uses cookies

Read our Privacy policy and Terms of use for more information.

We recently came across this great prompt guide from Alex Cooper of AdCrate, and it hit on something worth paying attention to: AI image generation is getting very fast, very practical, and very hard to ignore.

Cooper’s workflow shows how one Claude chat can be used to build a full brand DNA document, generate dozens of static ad prompts, and feed those into image tools like Nano Banana 2 or GPT Image 2 to create ad concepts at scale.

In plain English: the creative assembly line just got a lot shorter.

Here's exactly how this works:

  1. Open a chat in Claude and enter the Master Brand Research Prompt

  2. Download product images

  3. Enter the Claude Follow-up Prompt and Template Prompts into Claude to generate the and attach the product images

  4. Pick the templates you want to generate and enter them into Higgsfield

** I also want to give a shout out to Will Sartorius. He made the original Claude prompt, master brand research prompt and follow-up prompt. I just developed this into our own playbook with a swipe file of our own ads.

Step 1: Generate your brand DNA document

I created a Nano Banana 2 prompt document that will include everything you're going to need to create ads with this model.

First, we want to generate a comprehensive brand DNA document that Claude can use to create specific prompts for the image ads you’re going to make.

Start by opening a new Claude chat and dropping in the master brand research prompt (listed inside the prompt doc).

You only need to adjust two things in this prompt:

  • Brand name

  • Brand URL

That's literally it. Claude handles the rest, analyzing your brand and creating the context it needs for ad generation.

Step 2: Create brand-specific prompts

Once Claude generates your brand DNA doc, input the Claude follow-up prompt:

You only need to change three things in the follow-up prompt:

  1. Add your product images, you can paste these from the website (1-3 images work fine).

  2. Brand name.

  3. Your template prompts. I’ve added 40 different static ad types to the Nano Banana 2 prompt doc based on winning ads from my swipe file. These templates work for any brand because they're standardised formats that Higgsfield can run easily.

Step 3: Generate your ads in Higgsfield/Nano Banana 2

Use the 40 statics in the Nano Banana 2 prompt doc as your starting point.

Then, take Claude's generated prompts and run them through Nano Banana 2 in Higgsfield or connect the Higgsfield MCP to create all your ads inside of your claude chat.

Sometimes the copy in the Claude prompts isn't that great as we've not given it any context, so you can easily hook this up to one of your other Claude projects so it writes copy in your tone.

For the Aloha protein bars for example, here’s are a few ads I made:

All of these ads were pretty much ready to go straight into the ad account.

Once you have your workflow down using this system, you could get a VA to generate all the prompts in Claude and queue them in Higgsfield, or build this system out in Claude code like this guy did.

What blows my mind is how fast this stuff is moving. A few months ago, creating ads like this required designers, copywriters, and significant time investment. Now you can generate professional-quality static ads in minutes with almost zero manual work and with the right creative thinking baked in upfront.

Reply

Avatar

or to participate

Keep Reading