Back to Pragma Redact

Automated PDF, text & image anonymization. Powered by AI & smart patterns — 100% local.

No need to manually search and redact documents by hand. Pragma Redact automatically detects sensitive information (names, addresses, ID numbers, bank accounts, emails, and more) by combining AI with smart pattern rules. Mask everything in one click — 100% on your device, with no files uploaded to Internet!

Model not loaded

The first run downloads the personal-information detection model, so it's saved for next time. A recent version of a modern browser works best.

PDF anonymizer

Automatically detect and mask personal data across every page of your PDF document.

PDF input

Drop a PDF here or use the Upload PDF button.

—

Sample PDFs

Anonymized PDF

Upload a PDF to preview and anonymize the document.

Tip: drag a mask to move it. Drag its corner to resize. Use the trash button to remove it.

Text anonymizer

Paste any text — AI and smart pattern rules instantly spot and redact sensitive information.

Input text

Demo texts

Results

Labeled source

Waiting for detection.

Redacted text

Waiting for detection.

Image (OCR) anonymizer

Automatically extract text from images (OCR) and identify sensitive data without manual effort.

Image input

Image preview

Demo images

Result

Choose a demo image or upload your own image.

Tip: drag a mask to move it. Drag its corner to resize. Use the trash button to remove it.

Model parameters
0.20
Sets how confident the model must be before it flags something. A lower value detects more potential personal data (but may produce more false positives); a higher value only flags the most certain matches.
Smart pattern rulesPrecise pattern rules that complement AI to instantly catch exact-format identifiers (such as bank accounts, ID numbers, cards, emails, IP addresses). Toggle what you need or add custom rules — your choices are saved on this device.
Add a custom rule
Custom pattern rules let you detect organization-specific identifiers (e.g. ticket IDs, client numbers, internal codes).
Human-readable label displayed in the rules list.
JavaScript regular expression matching sensitive data (without surrounding slashes /).
Match modifiers: 'g' (all matches), 'i' (case-insensitive). Leave blank for default 'g'.