Images and PDFs
Scan PNG, JPEG, WebP, and non-encrypted PDFs up to 25 MB; PDFs can contain up to 20 pages.
Automatic PII redaction tool
Find likely PII, then decide what to redact.
Use automatic suggestions to speed up a careful privacy review without handing the final decision to a model.
What this tool does
Automatic PII detection works best as a second set of eyes
Personally identifiable information, or PII, is broader than a single list of patterns. An email address is easy to recognize. A person's name, an internal account ID, a handwritten address, or a face in a background photo requires more context and can be less certain.
FreeRedact combines optical character recognition, rule-based pattern checks, local name detection, visual face detection, handwriting-region detection, and custom text matching. Findings appear as suggestions with the visible value and location so you can decide whether each item should be covered.
Nothing is automatically removed merely because it was detected. Accept a suggestion, reject a false positive, redact every repeated match, or draw a manual box. This review-first design matters because an automatic system cannot know what the recipient is allowed to see or why a document is being shared.
Files and output
Supported formats and results
Text and visual detection
Review pattern matches, likely names and addresses, faces, and possible handwriting regions.
Custom repeat search
Enter a known name, project, customer, or identifier to find visible repetitions across the document.
Permanent reviewed output
Export only the boxes you approve into a new PNG or flattened image-based PDF.
How to use FreeRedact
A careful three-step workflow
Let the local scan finish
Open the file and allow the browser to render pages, read visible text, and create suggested findings.
Judge each finding in context
Read the detected value, inspect its location, accept private details, and reject harmless words or visual false positives.
Search, inspect, and export
Search for known repeated values, add manual boxes, perform a full visual pass, and download the reviewed result.
Automatic suggestions
PII categories FreeRedact can suggest
Different categories use different signals. A strong pattern match can be more reliable than a context-dependent name or face suggestion.
Text and visual categories
- Email addresses, common U.S. phone formats, payment cards, U.S. SSN-style numbers, and IPv4 addresses
- Person names, U.S.-style street addresses, account identifiers, sensitive URLs, and access tokens
- Faces and regions likely to contain handwriting
- Exact or case-insensitive custom text matches supplied by you
Why false positives and misses happen
- Names overlap with ordinary words, companies, job titles, locations, and headings
- Text recognition quality falls with blur, low contrast, tiny type, handwriting, rotation, and complex backgrounds
- A visual model can mistake logos, illustrations, or objects for faces and can miss partially hidden faces
- Unfamiliar secret formats and context-specific identifiers may require custom search or manual selection
Before you share
Human review checklist for automatic redaction
Measure success by the final file, not by how many suggestions the detector produces.
- Review every suggestion and reject false positives rather than accepting all findings at once.
- Search for known names, emails, account IDs, project names, and other values that repeat.
- Inspect headings, handwriting, tables, page edges, backgrounds, photos, and low-contrast regions manually.
- Add manual boxes for context-specific information the detector does not understand.
- Open the exported file and perform a second full review before sending it to the recipient.
- Rename the downloaded copy if its filename includes a person's name, case number, health condition, customer, project, or other private context.
Questions and answers
Frequently asked questions
What types of PII can FreeRedact detect?
It can suggest email addresses, common U.S. phone formats, names, U.S.-style street addresses, payment cards, U.S. SSN-style numbers, account identifiers, IPv4 addresses, private-looking links, access tokens, faces, handwriting regions, and matches in recognized text.
Does FreeRedact automatically apply every suggestion?
No. Findings are shown for review. You decide which suggestions to accept, which to reject, and where to add manual redactions.
Why can automatic PII detection miss a name?
Name detection depends on readable text and context. Decorative headings, uncommon names, OCR mistakes, handwritten text, and names that also resemble common words can reduce confidence or lead to a miss.
Is automatic redaction enough for legal or regulated documents?
Automatic suggestions can reduce review time but do not replace a qualified human review or your organization's legal and compliance process. Inspect every page and verify the exported file.
Related tools and guides
Continue your privacy check
Ready to review the file?
Open the private editor to apply these checks. No signup is required, and the file stays on your device.
Open private editor