Redact a name from a photo
Names are everywhere in the apps you screenshot — chat headers, profile cards, signatures, sender lines, sign-in prompts. Redacted runs a small named-entity model in your browser, finds names, and masks them.
Where names tend to hide
- Chat and email message headers
- Profile pictures with caption text underneath
- Order pages and account dashboards
- Signature lines in scanned documents
- Calendar invites and meeting screenshots
How to redact names automatically
- Drop or paste the image into Redacted.
- Click Detect text.
- The detector flags person names alongside other PII like emails, phones, and addresses.
- Review the masks; names that look like organisations or places sometimes need manual help.
- Download the cleaned image.
When the model needs help
The detector is biased towards common name shapes — typically two or three capitalised words. Edge cases:
- Names that share spelling with a famous brand or place ("Brooklyn Simmons" can read as a location)
- Single-word handles (e.g. "kirkov")
- All-caps names in screenshots
These cases are still safe to fix manually with a quick rectangle.
Privacy
The name detection model runs in your browser using a small int8 ONNX file bundled with the app. No images, no extracted text, and no embeddings ever leave the device.
FAQ
Will it pick up unusual or non-English names?
It tries. The model was trained on a mix of names but performs best on Latin-script names of typical length. Anything it misses can still be masked manually.
Will it accidentally mask brand or place names?
Sometimes — e.g. when a brand uses a person's name. Always review before sharing.
Can I keep some names visible?
Yes. Each detected mask is independent. Delete any masks you don't want and the underlying text will be visible in the export.