Redaction now covers images, not just documents. Scanned pages, screenshots, and photographs go through the same pipeline as a PDF: detect, review, redact, with the audit trail intact.
Text inside an image is found with OCR and passed to the same recognizers that handle documents, so a social security number is treated identically whether it was typed or photographed.
Features
- Redaction across PNG, JPEG, WebP, TIFF, and HEIC
- OCR for text inside images, including scans and screenshots
- Detection of visual identifiers — faces, signatures, and barcodes
- Blackout and blur strategies, applied per detection
- Page-level handling for multi-page TIFF
Improvements
- Low-confidence OCR is re-checked by a vision-language pass before it counts
- Redactions are burned into the output, so nothing is recoverable underneath
- EXIF metadata, including GPS coordinates, is stripped from output by default
Getting Started
Images work through the same endpoints and SDK calls as documents. See the API documentation for supported formats and detection types.