Toolpit
Loading your page…
Toolpit
Loading your page…
Delete words from the file, not draw a box over them.
No file to hand?(sample-invoice.pdf · made-up data)
Search for a name, a number or an address, choose the lines that hold it, and the instructions that draw those words are deleted from the page — not hidden under a rectangle that anyone can select, copy and read back. It works on the text of a page, and leaves annotations, form fields, document properties and attachments alone, so it suits removing a line from a document rather than sweeping every trace of a name out of a complicated one. The whole thing happens privately in your browser, which is where a document worth redacting should stay.
Drop in the PDF. Every page is read in this tab so the search can cover all of it.
Type the name, number or phrase you need gone. Every line containing it is listed.
Tick the lines to remove — or take all of them at once.
Save. The words are deleted from the file, and a bar is drawn where they were.
No. Toolpit does not upload your file for processing — the tool runs as code in this browser tab, works on the copy your browser opened, and forgets it when you close the tab. You can check this yourself: open your browser’s network panel and run the tool. You will see the page, its scripts and the ads on the site, and — when you save, if you are signed in — one small request that charges 10 tokens to your account. None of them carries your file, its name or anything read from it, because file processing stays on your device.
A black box is a shape drawn on top of the words, and both end up in the file. Select the page, copy it, paste it anywhere, and the text you thought you had removed is right there — which is how unredacted court filings and contracts keep reaching the press. This deletes the instructions that draw those glyphs from the page’s content stream, so on that page there is nothing left underneath the bar. The bar is cosmetic, drawn once the words are already gone. What this does not do is sweep the rest of the file: annotations, form field values, document properties, attachments and text inside reusable drawing objects are left as they are. Deleting a line of body text is exactly what this is for; clearing every trace of a name from a complicated document is more than it claims.
Not on its own. The smallest thing a PDF lets anyone delete is a run of glyphs the document drew as a unit, which is usually a line or part of one. If a line holds both something to remove and something to keep, redact the line and retype what should stay with the PDF editor.
No. A scan is an image, so there is no text in it to delete — what looks like words is pixels, and this tool works on text. Do not reach for OCR here: reading a scan produces a separate text layer and leaves every original pixel untouched, so removing a word from that layer hides it from a search while the picture of it is still plainly visible on the page. That is worse than not redacting, because it looks done. To redact a scan you have to destroy the pixels — cover the area in an image editor and re-export the page, then check the result by looking at it.
Open the saved file, select the whole page and paste it into a plain text editor. If the words still appear, the redaction plainly failed. Be careful with the opposite result, though: a clean paste shows that your viewer could not extract the text, which is weaker than showing the text is gone. Copying skips several places a word can still sit — annotations and comments, form field values, document properties, an attached file, or text drawn inside a reusable object. This tool removes the drawing instructions from the page’s own content stream and does not touch those other places. For a document where exposure would be serious, treat the paste test as one check among several rather than a clearance.
Toolpit adds no watermark or paid page quota. File and tool limits depend on device memory and supported features. Processing and previews are free; saving the result costs 10 tokens. A new account gets 50 tokens once its email is confirmed (once per email address) — five saves. Tokens pay for the download, never for a better version of it: the file you save is the same full-quality result either way.
Every one of these runs in your browser too — no upload, no watermark.