Skip to main content
The Image (Ad Generator) creates brand-new ad images from scratch. You provide your offer and brand, choose some creative options, and the AI generates multiple unique ads ready to download or upload to Meta. It does not copy anything, it generates original creative for your business. Where: sidebar → Image. Cost: 1 credit per variation. Plans: all. Illustration of generating original ad images from your brand and offer

Generate ads

1

Pick your brand

Use the brand switcher or the form’s brand selector. This fills in your business name, industry, logo, and colors. The Generator also reads the parts of that brand’s knowledge base that match your offer, so facts you wrote down there can show up in the ad.
2

Set the basics

Confirm your business name and industry (required), add a location if relevant, and choose an offer such as “Free estimate”, “50% off first order”, or a custom one.
3

Add visuals (optional)

Upload your logo (up to 2 MB) and up to five reference images (up to 10 MB each) to guide the style.
4

Choose format and variations

Pick an output format, a resolution (1K, 2K, or 4K), and the number of variations (1 to 4).
5

Pick an image model

The model chip beside the prompt picks which AI makes the image. The default is Nano Banana 2, and every model costs the same, so it is worth trying a different one when a result is not landing. See Image models below.
6

Fine-tune (optional)

Open Advanced options to pick ad types, psychology angles, a tone, and an objective. Leave blank to let the AI choose. See the options reference for the full lists.
7

Generate

Click Generate ads. The button frees up right away and results stream in as batches finish.

Image models

Every account can choose which image model makes the ad. Five are available today:
Every model costs 1 credit per image on every plan. Switching models never changes the price. The choice applies to that generation (or that edit) only; your default stays Nano Banana 2 until you change it in the picker.
The Ad Copier has no model picker. It uses whichever model Raddivus has set, so Copier results stay consistent for everybody.

What happens after you click Generate

Generation runs in the background, so you can navigate away or even close the tab. You will see placeholder cards immediately, and finished images replace them as they complete, usually within 30 to 120 seconds depending on how many you requested. If you reopen the page, in-progress batches resume automatically. Each card shows what is really happening. First it shows the step the run is on (Writing the ad, then Drawing the image) over a softly moving dot pattern. On the OpenAI models (GPT Image 2, 2.5 Flare and 2.5 Sunburst) you then watch the image take shape: a blurred preview with a percentage in the corner (25%, 50%, 75%), and the finished ad replaces it in the same spot. The number only moves when OpenAI sends a real preview, so a quick image can jump straight to the finished ad. A preview cannot be saved or downloaded, and previews cost no extra credits. The Google models (Nano Banana 2 and Nano Banana Pro) send nothing until the image is done, so their cards show the step until the ad arrives.
The number of images is set by the Variations field, not by how many ad types or angles you select. Three variations costs 3 credits, regardless of the other choices.

Working with results

Hover any finished image and tick its checkbox to select it. Select as many as you like, then use the selection bar to:
  • Save them to a folder
  • Download them (one image, or a ZIP for several)
  • Share them to the Ad Library
  • Delete them
  • Edit them with a follow-up prompt
Selecting, downloading, deleting, saving, and sharing are free. An edit re-runs the generator at the normal cost (1 credit per image).

Projects

Group runs into a project with its own instructions: pick one from the Project chip before you generate, or leave it on No project.

Advanced canvas view

The Ad Generator also has an Advanced canvas mode that shows your brand, style, and generated ads as a visual node graph. Every input, result, and action works the same as the basic view, just laid out spatially.

Troubleshooting

Large requests take longer. Total images equals your variations count, so reducing variations speeds things up. You can leave the page while it runs.
Use JPG, PNG, WebP, or GIF. The logo must be 2 MB or less, and reference images 10 MB or less each.
Make sure the brand color and typography toggles are on and that your brand has colors set in its profile.
Last modified on October 2, 2026