What PII gets flagged?

Names, addresses, dates, phone numbers, emails, passport and driver's license numbers, SSNs, bank account, IBAN and card numbers, PINs, usernames, URLs, IP addresses and VINs — Amazon Comprehend's PII entity types, kept deliberately high-recall because pruning a false positive in the editor is cheap. If Comprehend is unavailable, a regex fallback still catches emails, phone numbers, IBANs, US-format SSNs and card numbers.

Will redaction survive someone trying to lift the original text?

Solid-block redaction at export replaces the pixels entirely — there's no underlying text to recover. PDF redaction tools often fail at this (they hide text without removing it). We're working with raster images, which makes this honest.

Does the OCR text leave my session?

The OCR call goes to Google Cloud Vision, which returns the text to us, and the text is then sent to Amazon Comprehend for PII detection. We don't keep the text beyond the request lifetime.

Not directly — Blurify takes images (JPEG, PNG, WebP, HEIC/HEIF). Photograph or scan the page — or screenshot the PDF page — then redact it here and share the exported image. PDF support is on the roadmap.

Is an ID or passport photo safe to process here?

To read the text, the image goes to Google Cloud Vision for OCR — the same vision API every document here uses — and is auto-deleted within an hour; beyond that one call it reaches no other third party. The export strips metadata, and for the highest-stakes documents, Solid fill is the style that leaves nothing to reconstruct.

Redact a document online — forms, scans, IDs

Open the editor How it works

Desk photo of a laptop and a monitor with everything on both screens blurred out before sharing — Blurify result — Sample: two screens blurred whole. On a document photo, Google Cloud Vision OCR finds the text and Amazon Comprehend flags the PII.

Desk photo of a laptop and a monitor with everything on both screens blurred out before sharing — original — Sample: two screens blurred whole. On a document photo, Google Cloud Vision OCR finds the text and Amazon Comprehend flags the PII.

How it works for documents.

Drop the document photo or scan
Booking confirmations, ID cards, invoices, letters, forms — a photo or a flatbed scan. Anything with text that shouldn't be public.
Text + PII patterns highlighted
Google Cloud Vision DOCUMENT_TEXT_DETECTION reads every word, then Amazon Comprehend's PII detection flags names, addresses, dates, phone numbers, emails, ID and passport numbers, and bank and card numbers — those regions are pre-selected for you.
Pick a redaction style — export
Pixelation is the default for flagged text — fine blocks that leave nothing legible. Or pick a solid black block (the journalism convention), gaussian, or a color overlay. EXIF stripped automatically.

Questions, honestly answered.

What PII gets flagged?: Names, addresses, dates, phone numbers, emails, passport and driver's license numbers, SSNs, bank account, IBAN and card numbers, PINs, usernames, URLs, IP addresses and VINs — Amazon Comprehend's PII entity types, kept deliberately high-recall because pruning a false positive in the editor is cheap. If Comprehend is unavailable, a regex fallback still catches emails, phone numbers, IBANs, US-format SSNs and card numbers.
Will redaction survive someone trying to lift the original text?: Solid-block redaction at export replaces the pixels entirely — there's no underlying text to recover. PDF redaction tools often fail at this (they hide text without removing it). We're working with raster images, which makes this honest.
Does the OCR text leave my session?: The OCR call goes to Google Cloud Vision, which returns the text to us, and the text is then sent to Amazon Comprehend for PII detection. We don't keep the text beyond the request lifetime.
Can I upload a PDF?: Not directly — Blurify takes images (JPEG, PNG, WebP, HEIC/HEIF). Photograph or scan the page — or screenshot the PDF page — then redact it here and share the exported image. PDF support is on the roadmap.
Is an ID or passport photo safe to process here?: To read the text, the image goes to Google Cloud Vision for OCR — the same vision API every document here uses — and is auto-deleted within an hour; beyond that one call it reaches no other third party. The export strips metadata, and for the highest-stakes documents, Solid fill is the style that leaves nothing to reconstruct.