Redact PII in an image online
Personally Identifiable Information leaks easily through screenshots: a name in a chat header, an order number in a receipt, an email signature you forgot was on screen. Redacted scans an image for PII and masks it for you, all in your browser.
What counts as PII here
- Direct identifiers: full names, email addresses, phone numbers
- Locators: street addresses, postcodes, zipcodes
- Financial: credit card numbers, IBANs, bank account references
- Government: Social Security numbers, tax IDs, driver licences, passport numbers
- Technical: API keys, access tokens, IP addresses
How to redact PII in an image
- Drop the image into Redacted, or paste it from the clipboard.
- Click Detect text. OCR runs on the image and the PII detector flags each match.
- Each detected item gets a mask. Review them and add manual masks for anything missed.
- Pick a mask style — black box for high-risk text, blur for casual context, pixelate for faces.
- Download the cleaned image.
What the detector does — and doesn't — catch
The detector combines deterministic regex patterns (for emails, phone numbers, postcodes, financial identifiers, and credentials) with a small on-device named-entity model that flags person names. It is good at structured PII and well-known name shapes. It is weaker on:
- Handwritten or stylised text the OCR cannot read cleanly
- Internal identifiers that look like normal text (e.g. project codes)
- Sensitive numbers without a recognisable label
Always review before sharing.
Privacy
OCR and PII detection run entirely in your browser. The image is never uploaded. There is no account and no tracking.
FAQ
Is the OCR sent to a third-party API?
No. Both the OCR model and the PII detector are bundled into the app and execute locally.
Will it detect names that aren't in a dictionary?
The model handles a wide range of names, including unusual spellings. It can also pick up names that get mislabelled as organisations when the context is clear.
Can I redact only some of the detected items?
Yes. Each mask is independent — keep, move, resize, or delete any of them.