# AI images (https://sellersheetai.com/docs/mcp-api/workflows/images)

> Generate and edit Amazon-ready product images. Every delivered image lands on SellerSheet's CDN and in your store's Image Library, versioned per SKU and slot.



<Callout title="In short">
  Describe the image, or hand your agent reference images to edit. Jobs run in the background; the finished image comes back as a CDN link and is filed in your Image Library under its store, SKU and slot — nothing is written to Google Drive.
</Callout>

## How it works [#how-it-works]

<Steps>
  <Step>
    ### Generate or edit [#generate-or-edit]

    `generate_image` makes a new image from a description; `edit_image` changes an existing one — swap your product into a scene, recolour a variant, clean up a background. `reverse_prompt` turns a reference shot into a description you can reuse.
  </Step>

  <Step>
    ### Wait for the job [#wait-for-the-job]

    Images take a while, so the call starts a job. `check_image_job` (or `check_image_jobs` for several) reports when it is done and returns the image's CDN link.
  </Step>

  <Step>
    ### Filed in your library, versioned [#filed-in-your-library-versioned]

    Every generate and edit call names the store, the SKU and the slot, so the result is filed in the right place; each new image of a slot becomes the next version. `list_generated_images` lists what is there; the Image Library in the app shows it all.
  </Step>

  <Step>
    ### Disclose AI people [#disclose-ai-people]

    When an image shows a photorealistic person who does not exist, it carries Amazon's synthetic-performer marking. `set_synthetic_performer` changes the marking on an image after the fact.
  </Step>
</Steps>

<Callout title="Paid from Credits">
  Images are paid from your [Credits](/help/app/settings/billing). The price is shown and held before the job runs, and refunded when a job fails — a failed image costs nothing. Your store ref is checked before anything is held.
</Callout>

## Try asking [#try-asking]

> "Generate a white-background main image for SKU-A in myStore-US — product on a subtle shadow, no props, compliant with Amazon's main-image rules."

> "Make a white-background main image of my product in the style of this competitor shot."

> "Swap my product into this picture and keep the scene."

## Step 1, both ways [#step-1-both-ways]

The same tool and the same arguments — through your AI client, or from your own code.

**MCP** — you ask for the main image above. Your agent calls `generate_image` with the store, SKU and slot, polls `check_image_job`, and shows you the result for review.

<ToolCall tool="generate_image" args="{ prompt: &#x22;White-background main image of the product on a subtle shadow, no props&#x22;, sku: &#x22;SKU-A&#x22;, slot: &#x22;s0&#x22; }" />

<Callout type="warn" title="Write access required">
  Generating and editing create new images in your library and spend Credits, so the store needs write access (see [Stay in control](/docs/mcp-api/control)). Listing images and checking jobs only read.
</Callout>

## Tools on this page [#tools-on-this-page]

Pulled from the catalog at build time, each with its own access label.

<ToolList domains="[&#x22;image_gen&#x22;]" />

## FAQ [#faq]

### Do the images go to my listing automatically? [#do-the-images-go-to-my-listing-automatically]

No — they wait in your Image Library for review. Put them on a listing through [Create listings](/docs/mcp-api/workflows/listings-create) or [A+ Content](/docs/mcp-api/workflows/aplus).

### Is there a skill for this? [#is-there-a-skill-for-this]

Yes — the `image-gen` skill teaches your agent the compliance rules and the review loop. See [Install the skills](/docs/start/skills).