Create Amazon Product Images with AI and Publish Direct

Product images are the single biggest driver of click-through rate on Amazon, and most sellers know it. What they also know is how painful the process is: brief a photographer, wait days for product photography drafts, go back and forth on revisions, then manually upload each file to the right slot in Seller Central. For a catalogue of fifty products, that workflow can cost thousands of pounds and weeks of calendar time.

AI image generation has made the first part faster, but until recently the output still lived in a separate tool from the listing. You would generate an image in one place, download it, rename it to match Amazon's naming convention, and upload it through another interface. The generation was quick. Everything around it was not.

Rufusly's Image Studio handles the full loop: generate the image, add text callouts that are actually spelled correctly, optionally edit in Canva, and publish straight to your Amazon listing without leaving the browser. This article walks through how each piece works and where AI still has limitations worth knowing about.

Eight image types, generated from your listing data

The Image Studio generates eight types of product image, each sized and composed for a specific slot in an Amazon listing:

• Hero (main image): white background, product centred, sized for Amazon's main image requirements.
• Lifestyle: the product photographed in a realistic setting that matches its use case.
• Infographic: an infographic creator that places feature callouts over a styled background. The callout text is composited as a separate layer, not drawn by the AI model (more on why below).
• Detail grid: multiple close-up angles of the product in a single frame.
• Size guide: measurements and scale reference.
• Secondary lifestyle: a second lifestyle setting, useful for showing the product in a different context.
• A+ desktop: wide-format banner for A+ Content modules.
• A+ mobile: square format for A+ Content on mobile devices.

You can generate all six core types (everything except the two A+ formats) in one go by selecting "All Types", or pick a single type when you only need to fill one slot.

The generation is not blind. When you link an image to a product in your catalogue, the Image Studio reads the product's title, bullet points, description, and photos. The AI uses that context to generate images that reflect what the product actually is, rather than producing generic stock-style output. If you import product photos from Amazon (by entering an ASIN), those photos become source material for the generation.

Why callout text is composited, not AI-drawn

Rufusly composites callout text as editable layers rather than asking the AI to draw it, because image generation models misspell words. If you have tried AI-generated infographics, you have probably seen the problem: the model invents characters, renders text that looks almost right but is subtly garbled, or puts words in the wrong place. They approximate letter shapes rather than understanding text, and the approximation fails often enough that you cannot trust it for product claims on a live listing.

Instead, the AI generates the image without any text. Then a separate step reads your product listing and suggests callout labels (using a language model, not an image model, so the spelling is reliable). Those labels are placed as editable overlay layers on top of the generated image.

The placement uses a vision model to find the feature anchor points in the image, then positions each label in the nearest clear margin with a leader-line arrow pointing back to the feature. It auto-detects whether the margin is light or dark and picks the text colour accordingly.

The result is that every word is spelled correctly, every label is editable (you can move it, change the wording, adjust the style), and you can apply a single style across all callouts with one click. For a deeper look at the text-on-images problem, see our guide to fixing misspelled text on AI-generated Amazon infographics.

Edit in Canva without re-uploading

AI-generated images are a starting point. Most sellers want to adjust colours, add brand elements, or tweak the composition before publishing, and Canva is where that happens.

The flow is straightforward: click "Edit in Canva" on any image card, make your changes in Canva's editor (using your own Canva account and credits), and the edited version imports back into Rufusly automatically when you are done. If you close the Canva tab by accident, a "Bring back from Canva" button fetches the latest export. Reopening a previously edited image returns to the same Canva design rather than creating a new one, so your edits are preserved.

This is not a replacement for Canva's editor. It is a handoff: Rufusly generates the base image and manages the listing connection, Canva handles the fine design work, and the result flows back without a manual download-and-reupload step.

Publish to Amazon without leaving the browser

Once an image looks right, you can push it to your Amazon listing directly. Rufusly calls the SP-API to update the right slot on the right ASIN. You assign the slot (main image, PT01, PT02, and so on) when you generate or at any point afterwards.

Single-image publish works from any image card. For batch updates, select multiple images in the Creative Library and use bulk publish. The system tracks progress and reports which images succeeded and which failed (usually because the ASIN or slot assignment is missing).

Download with Amazon-ready filenames

If you prefer to upload images through Seller Central or a flat file rather than using the direct publish, Rufusly names downloaded files in the format Amazon expects for bulk upload: ASIN.SLOTCODE.jpg. For example, B0EXAMPLE01.MAIN.jpg for a main image, or B0EXAMPLE01.PT01.jpg for the first secondary image.

You can download a single image or select multiple images and download them as a zip file with all filenames pre-set. This saves the tedious step of renaming files manually before uploading through Amazon's bulk image tool.

Linking images to products

Every generated image can be linked to a product in your catalogue. This link is what makes publishing, callout generation, and filename conventions work. There are three ways to establish it:

1. At generation time: enter an ASIN or select a product before generating. The link is set automatically.
2. From the product page: navigate to a product in the Products Hub and open the Images tab. Generating from here pre-fills the product context.
3. After the fact: reassign the ASIN and slot on any existing image. This is free and does not cost a credit.

What AI image generation is not good at (yet)

It is worth being honest about where AI images fall short, because overselling the capability wastes more of your time than underselling it.

• Exact product likeness. AI models generate plausible-looking products, not pixel-accurate reproductions of yours. For a hero image that must show your exact product, a photograph is still the better starting point. AI-generated lifestyle and infographic images work well because the product is shown in context, and small differences are less noticeable.
• Fine detail on packaging and labels. If your product has specific label text, regulatory marks, or intricate patterns, the AI will approximate them. Use the generated image as a base and composite the real label in Canva.
• Consistency across a set. Generating all six types at once produces a coherent set, but the product may look slightly different across images. This is improving with each model update, but it is not solved.

The practical sweet spot today is: AI for lifestyle, infographic, detail grid, and A+ images (where the product is shown in context and exact fidelity matters less), and photography for the main hero image (where customers zoom in and scrutinise).

Daily limits and credits

Image generation uses credits. Each generation costs one image credit, and credits refresh monthly. There is also a daily cap to prevent runaway costs:

• Free trial: 10 images per day
• Starter plan: 25 images per day
• Growth plan: 75 images per day
• Agency plan: 150 images per day

If you hit the daily cap, the credit is refunded automatically. Editing in Canva, downloading, reassigning slots, and publishing to Amazon do not cost credits.