Detect names, emails, phones, and addresses locally, then add reviewed black visual covers. Underlying PDF data may remain searchable or copyable.
Drop PDF here or click to open
Supports PDF files up to 100MB
Uses OpenAI's open-source 1.5B-parameter privacy-filter NER model to recognize 8 categories of personal data. Review every candidate.
The model downloads when missing from browser Cache Storage and can be reused while the cache remains. Clearing or evicting the cache requires another download.
Detection runs entirely in your browser via WebGPU (with WASM fallback). Your PDF and all detected entities never leave your device.
One scan covers names, addresses, emails, phone numbers, URLs, dates, account numbers, and secrets — far beyond what regex can do.
Toggle which PII types to detect with checkboxes above the panel. Your selection is remembered across sessions automatically.
Drag the threshold slider (0.50–0.99) to control how strict the AI should be — higher values mean fewer false positives.
Drag and drop or click to select the PDF you want to scan.
The ~900 MB privacy-filter model downloads when it is not cached. Scan time varies with device, browser, and document size.
Filter and review detected entities, then add visual black covers. The underlying PDF content is not guaranteed to be removed.
Smart Redact can scan a PDF for names, email addresses, phone numbers, locations, and other likely personal information. It downloads an OpenAI privacy model to the browser when needed, then runs detection on the device so the PDF is not uploaded for AI analysis.
Every match is only a candidate for you to review. Applying a match creates a visible cover and does not guarantee removal of the underlying text layer, so do not rely on it alone for confidential disclosure; verify that sensitive text cannot be searched, copied, or extracted from the exported file.
Yes. OnePDFs is free to use, requires no account, and does not add an extra OnePDFs branding watermark; watermark tools add only the watermark you choose.
Your files stay on your device and are processed locally in your browser; no upload to our PDF processing service is required. A browser copy can remain in this site's IndexedDB until you clear site data or the browser evicts it.
It uses OpenAI's open-source privacy-filter model — a 1.5B-parameter NER (named entity recognition) model trained to identify 8 types of personal information: names, addresses, emails, phone numbers, URLs, dates, account numbers, and secrets.
Accuracy varies by language, document layout, and data type. Review every candidate, test the model on representative documents, and use the confidence threshold to balance missed items against false positives.
Around 900 MB (q4 quantization). It downloads when missing from your browser's Cache Storage and can be reused while that cache remains. Clearing or evicting the cache requires another download.
No. The entire detection pipeline runs in your browser using WebGPU (or WASM as a fallback). Your PDF and all detected entities never leave your device.
You stay in control. After scanning, you can disable specific PII categories with one click, raise the confidence threshold, or remove individual entities before applying. Nothing is redacted until you click Apply.
No. Apply adds black visual covers; the underlying PDF text may remain searchable, copyable, or extractable. Use a content-removing redaction workflow for sensitive data and verify the exported file in another reader.