BleachPDF
github.com
- Free plan
- Open source
- macOS
- Linux
- Self-hosted
What BleachPDF is
BleachPDF is a free desktop tool for removing sensitive text from PDFs on Linux and macOS, including scanned pages. It turns each page into an image, uses OCR to locate specified text, covers matches with black boxes, then saves a new PDF. Since the result consists of images rather than a text layer, redacted wording cannot be exposed by copying and pasting. You can search for exact text or set patterns, including case-insensitive matches and digit patterns, and process a file or folder pattern from the command line. BleachPDF checks its output by scanning again; if a match is not hidden, verification fails. The README says documents stay on your computer and are not uploaded. Installation uses pip and requires Tesseract, with setup instructions for Ubuntu/Debian and macOS. You can choose OCR languages, with examples including eng and eng+kor. Recognition may be unreliable with handwriting, unusual fonts, poor scans, or tiny or dense text, so manual review is advised. The repository displays an AGPL-3.0 license and states that it grants no permission for censorship as defined in the license.
Who it suits
BleachPDF suits Linux and macOS users who need to redact text in PDF files, including scanned documents, and can review OCR results manually. It may also suit users who want command-line batch processing and local document handling.
Good to know, and what to watch
- Works with scanned PDFs using OCR.
- Processes a file or folder pattern.
- Checks output and fails on verification errors.
- Documents remain on the user's computer.
- Supports selectable OCR languages.
- Requires Tesseract installation.
- OCR can err on handwriting or poor scans.
- Manual review is advised.
- Not available for Windows.
Plans
Every plan and priceBleachPDF
Free plan
The verdict
BleachPDF offers a focused local workflow for PDF redaction, with output verification and batch processing. OCR limitations make checking the saved document important.
What the maker publishes
- Batch processing
- The command line accepts a folder or file pattern to redact multiple PDFs. source
- Batch processing
- The CLI accepts folders and glob patterns to process multiple PDF files. source
- Benchmark testing
- The project says its tests use Allen Institute for AI's olmOCR-bench and rescan outputs to check that redacted text is hidden. source
- Case-insensitive matching
- Pattern syntax supports case-insensitive literals using the i modifier. source
- Configuration files
- Configuration can be selected explicitly, through BLEACHPDF_CONFIG, or from user and system configuration locations. source
- Dependencies
- Installation uses pip and requires Tesseract; the README gives setup instructions for Ubuntu/Debian and macOS. source
- Dependencies
- Declared dependencies include pytesseract, Pillow, ReportLab, PyMuPDF, PyYAML, parsimonious, platformdirs, and psutil. source
- Developer interface
- The package exposes a bleachpdf console-script entry point calling bleachpdf:main. source
- Digit and letter patterns
- The pattern language can match arbitrary digits or letters, such as nine-digit or ten-digit sequences. source
- Exact matching
- Users can supply literal strings with the -m or --match command-line option. source
- Exit statuses
- The CLI documents distinct codes for configuration errors, file errors, no matches, and verification failures. source
- External benchmark details
- The olmOCR-bench dataset page identifies it as an Allen Institute for AI benchmark with 1K to 10K document examples and English-language data. source
- Image-only output
- The resulting PDF contains page images without a hidden text layer that could expose redacted text through copying. source
- Installation dependency
- Installation requires the Tesseract OCR engine, with Ubuntu/Debian and macOS setup commands documented. source
- Language support
- The OCR language can be selected with --lang, with examples including eng and eng+kor. source
- License
- The repository displays an AGPL-3.0 license and states that it grants no permission to use the software for censorship as defined in the license. source
- License
- The package is licensed as GNU AGPL-3.0-or-later and is classified as OSI-approved AGPLv3+. source
- Limitations
- Handwriting, unusual fonts, low-quality scans, and very small or dense text can cause recognition errors, so the README advises checking output manually. source
- Limit: censorship restriction
- The accompanying license states that permission is not granted to use the software for censorship as defined in the license text. source
- Limit: manual checking
- The author says users must inspect results themselves because automated redaction is not perfectly reliable and may over-redact. source
- Limit: OCR accuracy
- The README warns that handwriting, unusual fonts, poor scans, and very small or dense text can be misrecognized. source
- Local privacy
- The README says processing occurs on the user's computer without uploads, accounts, or subscriptions. source
- Maker
- The README identifies the author as John Byrd. source
- Maker platform claim
- John Byrd stated that the tool works on both vector PDFs and scanned PDFs. source
Checked against the maker's pages on 15 Oct 2026.