Carousel Ad Images with AI: Step-by-Step Workflow

To make carousel ad images with AI, connect Polaris to Claude once, then ask Claude in plain chat for each carousel frame (attaching your product photo), pick settings from the tappable option cards, and watch the live panel render each image before downloading the set as a ZIP.

What you need before starting

• A product photo, ideally a clean shot with good lighting, ready to attach in chat.
• A rough carousel story: a hook frame, two to four supporting frames, and a closing offer or CTA frame.
• The Polaris MCP connector added to Claude, which is a one-time setup covered below.

Step-by-step: generating your carousel

1. Add the connector. In Claude, add the Polaris connector at polarisugc.com/api/mcp and authorize it. This only needs to happen once; after that Polaris is available in any conversation.
2. Attach your product photo. Drop the image into the chat and tell Claude what the carousel is for, e.g. "make a 4-frame carousel ad for this water bottle, hook first."
3. Pick your settings from the cards. Claude surfaces tappable option cards for things like image model (Nano Banana Pro or GPT Image 2), aspect ratio, and style. Tap through instead of typing settings manually.
4. Watch the live panel. A Polaris panel appears directly in the conversation and renders progress, then the finished frame, usually within 20 to 180 seconds per image.
5. Request the rest of the set in parallel. Ask Claude for the remaining frames in the same message or a follow-up; Polaris can run them as a parallel batch instead of one at a time.
6. Download the ZIP. Once every frame in the batch finishes, download the whole set as a ZIP rather than saving images one by one.
7. Add captions if needed. If any frame needs on-image text, use the one-click burned-in captions option instead of editing frames externally.

Pro tips for stronger carousels

• Lead with the strongest visual, not the most detailed one — the first frame decides whether anyone swipes.
• Vary the composition across frames (close-up, in-use, lifestyle, offer) so the carousel doesn't feel like one image repeated four times.
• Use one-click Recreate on any frame that almost works instead of writing a brand-new prompt from scratch.
• Run Copy Judge on your on-image or caption copy before shipping the set, especially for the closing CTA frame.
• If a render fails, don't worry about wasted spend — failed renders are auto-refunded.

Honest caveats

Polaris builds carousel images around your actual product photo — there's no stock-avatar or stock-scene library standing in for it. Treat every generated set as a first draft to test, not a guaranteed winner; run different hooks and frame orders through real ad tests before scaling spend. If you're in a regulated niche (health, finance, supplements), keep your on-image claims compliant — avoid disease, cure, or guaranteed-result language, since Polaris won't fabricate those claims for you but also won't stop you from writing them yourself.

Where this fits

Carousel images are one output of the same workflow used for AI UGC video ads — same connector, same chat-based flow. If you want the fuller picture of how the Claude-based workflow works end to end, read this walkthrough of generating ads inside Claude via MCP. When you're ready to generate at volume, see pricing for credit details, or jump straight into the app if you'd rather work in a browser first.