How to Create Static Ads Like a Pro Using AI (Step-by-Step)
Wanderson Jackson
Running paid ads without strong creative is like lighting money on fire. The problem? Hiring designers is slow and expensive, and most AI tools still require you to be a prompt expert just to get something decent.
In this guide, I'll show you the exact workflow I use inside Avocado AI to go from a product photo to a high-converting static ad — fast, repeatable, and no design skills required.
Full breakdown of the AI Static Ads
What You'll Need
Before we start, grab these three things:
Avocado AI (avocadoai.co) — where you'll generate and organize your creatives
Claude AI — to build your JSON prompt templates (no prompt expertise needed)
Pinterest — for ad inspiration and layout references
That's it. No Photoshop. No agency. Let's get into it.
Step 1: Set Up Your Storyboard
Log into Avocado AI and head to Storyboards. Unlike Workspace (which is better for linear folder organization), Storyboards lets you visualize everything in one place — all your product images, templates, and generated creatives side by side.
Upload your product photos as reference images. A few tips here:
Upload multiple angles of the same product — the more the AI understands your product's shape, label, and packaging, the better it will maintain design consistency in generated ads.
Aim for the highest quality and clarity possible.
Mark each image with a status (Approved, Pending, Needs Review) to keep your workflow clean.
Step 3: Get Inspiration from Pinterest
Go to Pinterest and search for ad styles you like — ecom product ads, beauty campaigns, lifestyle layouts. When you find something you want to recreate, copy the image.
Other platforms like Behance and Dribbble work too, but Pinterest is the fastest for ad-specific inspiration.
Step 4: Use Claude to Generate a JSON Prompt Template
This is where the workflow gets powerful. Open Claude AI, paste your Pinterest inspiration image, and ask it:
"Create a JSON prompt template based on this image. I want to use this as a reusable template for generating static ad variations."
Claude will output a structured JSON prompt with editable variables — things like background color, brand color, icon style, copy, and layout format. JSON prompts work especially well for AI image generation because the structured format gives the model clearer instructions than a block of plain text.
Save this prompt. You'll reuse it across multiple products.
Step 5: Generate Your Static Ad
Back in Avocado AI:
Click on your product image
Hit Edit Image
Paste your JSON prompt into the prompt field
Select Nano Banana as your generation model
Set your aspect ratio (16:9 for landscape, 9:16 for portrait/stories)
Choose the number of variations and your resolution
Hit Generate
Review the output. Iterate if needed. Your settings are saved, so you don't have to redo them every time — just swap the prompt or product image and generate again.
Step 6: Upscale for Maximum Quality
If the resolution isn't quite where you want it, don't just download raw. Use the upscale feature inside Avocado AI before downloading — this sharpens the image significantly and makes your creatives look more professional, especially at large sizes.
Step 7: Repeat and Scale
This is where the real leverage is. Once you have one prompt template working:
Copy it
Swap in a different product image
Adjust a few variables (brand color, label text, background)
Generate
You're not just changing copy — you're testing entirely different layouts and visuals. This is how you build a real testing library for your ad campaigns without burning days in Figma.
What's Coming Next
Avocado AI is rolling out preset templates inspired by winning ads — so you won't need to manually build JSON prompts from scratch. Just pick a preset, drop in your product, and generate.
Brand DNA is also launching soon — it will automatically pull your brand colors, typography, and visual style and inject them into every prompt, so your ads stay on-brand without any manual configuration.
Try It Yourself
The full storyboard from this tutorial — including all prompt templates — is publicly available. Drop a comment below and I'll send you the link.
