API Documentation

Send a PDF and the watermark text you want removed. The API processes the document synchronously and returns the result in the same response, together with a machine-readable report that says exactly what was (or was not) removed.

Authentication

Optional. Anonymous requests are limited per client IP to 10 per minute and 60 per hour. An API key sent as X-API-Key gets its own, higher quota and attributes requests to your client name. Request a key via the contact link in the footer.

Limits

One PDF per request, application/pdf, up to 10MB when calling the API directly. The website uploader is limited to 10MB. Processing is synchronous and typically takes a few seconds; allow a timeout of at least 120 seconds.

What it removes

Text watermarks in the PDF text layer that match the text you supply. Scanned or image-only pages have no text layer and are reported as no_text_layer.

Uploads

Uploaded files are handled as described in the Privacy Policy. Do not send documents you are not permitted to modify.

A 200 means the request was processed, not that a watermark was removed. The response always contains a PDF; check outcome.status in X-Processing-Report to know whether anything changed.

Example X-Processing-Report

{
  "summary": {
    "text_watermarks_found": 9,
    "text_watermarks_removed": 9,
    "image_watermarks_found": 0,
    "image_watermarks_removed": 0,
    "patterns_matched": ["CONFIDENTIAL"]
  },
  "outcome": {
    "status": "removed",
    "reason": null,
    "discovery": {
      "status": "uncorroborated",
      "discovered_target": null,
      "similarity": 0.453,
      "adopted": false
    }
  },
  "pages": {
    "1": {
      "page": 1,
      "text_watermarks": 3,
      "image_watermarks": 0,
      "location_count": 1,
      "location_types": ["text"],
      "has_reported_change": true
    }
  },
  "warnings": [],
  "diagnostics": { "...": "engine internals, shape may change" }
}

• summary holds the run totals. pages has one entry per page the engine touched, and text_watermarks / image_watermarks count the occurrences removed on that page. A shared form drawn on several pages counts once per page, so the per-page sum can exceed the total. Treat has_reported_change as the per-page yes/no; a page listed only with chunk_text evidence changed but could not be counted. location_types values include text, image, and chunk_text.
• outcome.discovery is informational. found with adopted: true means the engine used a close variant of your text (discovered_target); no_text means no text layer. Other status values describe why no variant was adopted, and the set may grow.
• summary, outcome, pages, and warnings are stable. diagnostics exposes engine internals and may change between releases.

curl -sS -X POST https://api.watermarkremoval.com/api/v1/remove-watermark \
  -F "file=@input.pdf;type=application/pdf" \
  -F "watermark_text=CONFIDENTIAL" \
  -D headers.txt \
  -o output.pdf

# The verdict is in the X-Processing-Report header
grep -i '^x-processing-report' headers.txt

The OpenAPI document carries every header and error shape on this page, with stable operation ids (remove_watermark, remove_watermark_stats) for generated clients. https://api.watermarkremoval.com/health reports service status and the deployed engine_version.