Polaris
AI modelsNano Banana Pro for Ad Images
AI model

Nano Banana Pro for Ad Images

2K and 4K static ad images that keep your real product accurate, from 15 credits

Nano Banana Pro is the workhorse image model in Polaris: it generates static ad creative at 2K (15 credits) or 4K (22 credits), outputs PNG, and takes up to 8 reference images via image_input so the product in the shot is recognizably yours — same label, same shape, same color.

What Nano Banana Pro is best at

Product-faithful stills. Feeding several angles of your real product through image_input is what separates a usable ad image from a generic AI render: the model anchors to what your product actually looks like instead of hallucinating a lookalike. That makes it the pick for lifestyle scenes, before/after frames, feature callout graphics, and thumbnail stills to pair with the video ads Polaris generates.

At 15 credits for a 2K image, it is also the cheapest single render in the whole Polaris lineup — an order cheaper per asset than any video clip. Static creative still earns placements in most ecommerce accounts, and the low unit cost means you can batch dozens of concepts without denting the credit pool. See pricing for the math.

Specs at a glance

  • Type: image (static ad creative)
  • Credits: 2K = 15, 4K = 22
  • References: up to 8 images via image_input for product fidelity
  • Output: PNG

When to pick it over the video models

Reach for Nano Banana Pro when the placement calls for a still — carousel cards, catalog-style creative, static retargeting — or when you want to storyboard concepts cheaply before spending video credits. A common pattern: generate a batch of stills to find the scene and angle that feels right, then hand the winner to Kling or Veo as the basis for a moving version of the same idea within your UGC ad mix.

The trade-off is inherent: it is an image model. No motion, no sound, no talking actor. If the concept depends on demonstration or delivery, start with a video model instead. And as with everything generated, a clean render is a starting hypothesis — the market picks winners, not the model.

How to use it

  1. Connect the Polaris MCP once, then ask Claude in plain chat for the image, attaching your product photos as references.
  2. Pick Nano Banana Pro on the tappable model card and choose 2K or 4K.
  3. The live panel shows the finished PNG in the conversation; failed renders auto-refund.

Setup details are in how to generate ads inside Claude via MCP.

Frequently asked questions

How much do Nano Banana Pro images cost?
15 credits at 2K and 22 credits at 4K, output as PNG — the cheapest single render in the Polaris lineup, which makes batching many static concepts inexpensive.
How does Nano Banana Pro keep my product accurate?
It accepts up to 8 reference images via image_input. Feeding multiple angles of your real product anchors the render to your actual label, shape, and colors instead of a generic lookalike.
When should I use an image model instead of a video model?
For static placements like carousels and retargeting, or to storyboard concepts cheaply before spending video credits. If the ad depends on motion, sound, or a spoken script, start with Kling, Veo, Seedance, or Omni instead.
What file format does Nano Banana Pro output?
PNG, at 2K or 4K resolution, delivered in the live Polaris panel inside your Claude conversation.

Generate winning ads inside Claude

Connect Polaris to Claude and generate UGC-style video and image ads right in the chat. First batch of 12 free.

AI models

GPT Image 2 for Ad ImagesKling 3.0 for AdsOmni Talking-Actor ModelSeedance 2.0 for Ads

Related reading

GuideGenerate ecommerce ads inside Claude (MCP guide)GuideImage to video: turn a product photo into a video ad