Redact PDF

Names, numbers and selected areas are deleted from the file, not hidden under a black rectangle, and a report proves it. All of it in your browser.

Open in PDF ARENA

It removes, it does not cover

The redacted page is rasterised. No text object survives underneath.

It finds it for you

Emails, phones, ID numbers, IBANs and cards, checked with their own arithmetic.

It proves it

The result is reopened and searched. You get the report with both hashes.

Redaction that removes, not covers

A black rectangle drawn over a name is not redaction. It is a drawing. Open that file in another viewer, select the area, copy, paste: the name comes straight out, because the object underneath was never touched. Newsrooms and law firms have leaked documents exactly that way. Here it runs backwards. Apply a redaction and the page carrying it gets rasterised — a picture of itself, your areas painted solid. Nothing survives below the black. No glyph survives on that sheet at all. Pages you never marked are copied verbatim, so a fifty-page contract with two censored lines keeps its vectors, links and selectable words everywhere else. It all happens inside your browser. Nothing is uploaded.

How to redact a PDF

1

Open your PDF

Drop the file in. It stays on your machine, and once this page has loaded the tool keeps working with the network off.

2

Review what was found

Emails, telephones, identity codes, IBANs and cards arrive with a confidence level and the reason behind each. You pick.

3

Add whatever else

Hunt a word or a regular expression document-wide, drag a box over a signature, or mark a whole sheet.

4

Apply and read the report

The output is reopened, its text pulled back out, every hidden value searched for. You watch the checks and download a report carrying both fingerprints.

What gets removed, and how it is checked

Detection reads the text as the file stores it, never a flattened dump. Fragments get stitched back into reading lines, so a telephone split across three of them still surfaces while one table column stops gluing itself to its neighbour, and only then does arithmetic rule: the control letter of a DNI, modulo 97 over an IBAN, Luhn across a card. Whatever fails the maths arrives as a mere pattern, lower confidence, labelled. Nothing applies behind your back. Applying runs in three stages. Order matters. Every marked sheet is rasterised and its boxes painted; then nine points inside each box are read back to confirm those pixels went black, because a canvas beyond the browser's size ceiling accepts drawing commands and quietly paints nothing, and that silence is precisely how censorship ends up cosmetic. Writing begins only once all sheets pass. Metadata, annotations, attachments, embedded JavaScript and launch actions then go, following your switches. The last stage turns a claim into a fact. The finished PDF is reopened through a separate code path, its text extracted, each hidden value hunted again with separators and casing stripped. Rasterised sheets are checked for returning zero characters. Metadata, for being bare. Should anything fail, the job is declared failed and no download appears. The report holds the SHA-256 of input and output, every check with its verdict, and each removed value as a fingerprint plus its final two characters: enough to recognise, useless for rebuilding.

Why this one

Everything executes inside your browser. The document goes to nobody.

DNI, NIE, IBAN and card numbers validated by their own arithmetic, not guessed from a shape.

Each job closes with a verification pass, and a failed check means no download.

Questions about redacting PDFs

Short answers, limits included

Can the redacted text be recovered?
Not from the sheets that carry them. Those get rasterised, so the glyphs are gone and nothing sits beneath the black. It is checked, not assumed: each job reopens the finished PDF, extracts its text and hunts every removed value. If one surfaces, no file reaches you.
Does the whole document turn into an image?
Only the sheets you marked. The others are copied as they stood, keeping vectors, links and selectable words.
What happens to metadata and attachments?
By default title, author, subject, keywords and the XMP packet are emptied, while attachments, embedded scripts, launch actions and form definitions go. Each is a switch you can flip back.
Does it work on scanned PDFs?
Boxes work: you drag, those pixels vanish for good. Automatic proposals need a text layer, so a pure photograph yields none. Run OCR first.
Is anything sent to a server?
Nothing. Detection, removal and verification all execute in your browser through pdf.js and pdf-lib. The report is assembled on your machine too, and what you download never left it.

Updated on

Do more with your PDFs

Discover all the professional tools we have prepared for you. All in one place.