Open DevTools.
Black out a name, a number, a paragraph.
Watch nothing get sent anywhere.
Draw a rectangle in page coordinates and ClientPDF removes the text runs underneath it from the file's own content stream, then draws a black box over the area — entirely inside this tab. Zero network requests after the page loads, same rule as every other tool here. Check the Network tab yourself.
Drop one PDF here or click to browse
One file at a time · stays on this device
Honest note: only text drawn with the Tj/TJ text-show operators is removed from the file — a solid black box is drawn over every area regardless, but images, scans, and vector graphics under the box are only visually covered, not deleted. Bounding-box matching assumes roughly horizontal, non-rotated text; rotated text may be under- or over-matched. Coordinates are in PDF points, origin at each page's bottom-left corner — use the page size shown above once a page number is entered.
Why trust this tool
No accounts
No signup, no email, no watermark. Use it and close the tab.
No limits (for now)
Queue as many redaction areas, across as many pages, as you need. No artificial caps.
Cropping margins or deleting whole pages too? Both live too.
Trim a page's visible margins or remove entire pages — same rule: everything stays local. Leave your email and we'll let you know about a Pro batch mode.
Thanks — saved locally for now, we'll be in touch once signups are live.
Removing the text, not just hiding it
Redact tokenizes each affected page's own content stream — a real parser that tracks the text matrix, current transformation matrix, font size, and line leading operator by operator — and drops every Tj, TJ, ', and " text-show operator whose estimated bounding box overlaps a rectangle you drew. Matched text is deleted from the stream before the file is saved; it isn't merely covered. A solid black rectangle is then drawn over the same area for every match, so non-text content — images, scans, vector art — is also visually covered even though it isn't removed from the file.
Worth knowing before you rely on this for anything sensitive: the bounding-box estimate is a deliberate overestimate based on average character width for the active font size, tuned to remove slightly too much text rather than leave a partial word behind — it assumes roughly horizontal, non-rotated text, so rotated or heavily skewed text may not line up with your rectangle. And only Tj/TJ-drawn text is actually deleted; an image of a document (a scan) has no such text to find, so the black box is the only protection you get there. If you just need to remove entire pages rather than part of one, Delete Pages is the more direct tool, and if you only need to hide a margin without touching content, Crop does that — though as that page notes, cropping doesn't delete anything either.