Redact a PDF

Redact PDF removes the parts of a document you mark, rather than simply covering them. You open a PDF, draw a black box over each name, account number, signature or paragraph that must not leave your hands, and download a copy in which that content no longer exists. It is for sending a contract, a payslip, a medical letter or a support ticket to someone who should see the rest of the page but not the private part.

The most important thing to understand is why a black rectangle is not enough. If you draw a filled black box on an ordinary text PDF — in a viewer, or by adding a shape with an editor — the words underneath are still there. They are still text, still selectable, still found by search, and still recoverable by anyone who opens the file with the right tool and reads the objects below the box. That is cover-up, not redaction, and it is how redactions have leaked in real life. To genuinely remove the words, the page has to stop being text.

This tool does that. Each page you draw a box on is rendered into a picture at the resolution you choose, your boxes are painted into that picture as solid black, and the page in the downloaded PDF is that picture instead of the original text. The covered characters are not hidden behind a box — they are not in the file at all after the page is rebuilt, because the rebuilt page contains only pixels. The cost is exactly what you would expect: a rebuilt page is an image, so its text can no longer be selected, searched, copied or read by a screen reader, and the file is usually larger than before. The pages you mark keep their size and their layout, because the picture is the same page drawn at the resolution you chose; only their nature changes, from text to image.

Just as important is being precise about which pages change, because a tool that quietly rebuilds a whole document would be as surprising as one that quietly fails to. Only the pages you actually draw a box on are rebuilt. Every other page — the ones you never opened, and the ones you opened but did not mark — is copied through completely unchanged, so their text stays selectable and searchable and their file size does not grow. If you redact page 3 of a twelve-page report, pages 1, 2 and 4–12 come through as they were. The result panel tells you how many pages were rebuilt, and the box count on each page is shown as you work.

The resolution you pick sets where the boundary between quality and size falls. At 72 DPI a page is roughly screen resolution, good for a quick job; 150 DPI is the default and reads comfortably on screen; 300 DPI keeps fine print crisp for printing. A single rebuilt page cannot exceed 100 megapixels or 16,384 pixels per side, and a whole job cannot exceed 1,000 megapixels, because everything is held in your browser's memory — when a request goes over, the tool names the page and suggests a lower resolution. Password-protected PDFs are refused and have to be unlocked and re-saved first, and files up to 200 MB are accepted.

Two things this tool does not do, so you know before relying on it. It does not remove hidden data that is not drawn on a marked page — if your private text sits on a page you did not mark, that page is copied through with the text intact, so mark every page that carries something private. And because the output is a new document assembled from the pages, document-level items that live outside the pages are not carried over: bookmarks, form fields and the document title/author metadata are dropped, and a link that pointed at a rebuilt page is removed. Links between pages you did not redact keep working.

Your PDF never leaves your device. It is read and rebuilt inside your browser tab with pdf.js and pdf-lib; you can confirm in the Network tab of your browser's developer tools that no request carries the file.

1. The file's first bytes are checked for the %PDF- header, so an image renamed to .pdf is named for what it really is. Files over 200 MB are refused before they are read into memory. 2. pdf.js opens the PDF and reports each page's displayed size in PDF units (1/72 inch), with any page rotation applied. A password-protected or damaged file is refused with a message that says why. 3. The page you are looking at is drawn to a canvas at preview resolution — a scale of min(2, 1000 ÷ page width, 1600 ÷ page height). You draw boxes on that canvas; each box is stored in preview pixels with a top-left origin. 4. For each marked page, every box is converted twice: preview pixels → PDF points, flipping the vertical origin because PDF y grows upwards from the bottom-left, then PDF points → output-image pixels with a top-left origin, at the chosen DPI. Edges are rounded outwards (floor the left and top, ceil the right and bottom) so a box never leaves a one-pixel sliver of the content it covers. 5. The resolution is checked first: pixel size = points × DPI ÷ 72. A rebuilt page over 100 megapixels or 16,384 px per side, or a job over 1,000 megapixels in total, is refused before anything is drawn, naming the page and suggesting a lower resolution. 6. Each marked page is rendered again at the chosen DPI, the boxes are filled solid black onto that image, and the image is encoded as a JPEG (quality 90, on white). Pages with no box are copied through byte-for-byte. 7. The output is assembled in page order — copied pages and rebuilt pages interleaved — and saved as <name>-redacted.pdf.