Word-Level Bounding Boxes

Standard extraction returns block-level bounding boxes for text, tables, and figures. Turn on word-level bounding boxes when you need exact word coordinates for review overlays, field-level QA, redaction previews, evidence packs, or high-trust citation UX. Platform sample: Bank Statement

Word-level bounding box toggle in the Platform

{
  "file_url": "https://platform.runpulse.com/api/examples/637e5678-30b1-45fa-acc4-877f2d636419/pdf",
  "extensions": {
    "alt_outputs": {
      "wlbb": true
    }
  }
}

Word-level output is returned at extensions.altOutputs.wlbb. It can make results much larger, so enable it for review and QA workflows rather than every ingestion job. The equivalent top-level word_level_bounding_boxes: true parameter is also accepted. Note that the standard bounding_boxes.Words collection — word text, page, confidence, and polygon — is always part of the extraction response; extensions.alt_outputs.wlbb adds the separate word-box output used by existing WLBB workflows.

If your workflows always need word-level output, your organization can have it enabled by default on every extraction without passing the parameter per request — contact support to configure an org-wide default.

Bank statement with word-level bounding boxes overlaid

For most review workflows, the standard bounding box layer is enough:

Bounding Boxes

Understand block-level and word-level source coordinates.

⌘I

Start Here

Agent Tools

Core Concepts

Security & Compliance

Help

Word-Level Bounding Boxes

Bounding Boxes