reloadium.com

annotateImage

Detect and label the distinct objects/areas in an image, and show it in chat with numbered boxes drawn over each one (labels are listed below the image, not drawn on it, to avoid clutter). This is a cheap vision call, not the image-generation model — not a billed generation, so it's fine to call before writing a precise editImage instruction, e.g. when the user says 'remove the thing on the left' and you don't yet know exactly what's there or where. Pass the returned annotations array straight into editImage's `annotations` param to ground that edit to the right region. By default annotates the most recently generated/edited image in this conversation; pass imageUrl to annotate a real photo from the web instead (e.g. one found via getBestImage).

Page: /

Input schema

{
  "type": "object",
  "properties": {
    "imageUrl": {
      "type": "string",
      "description": "Optional: a public image URL to annotate instead of the most recently generated/edited image."
    }
  }
}