Redact a PDF offline, without uploading it.
Your file never leaves your device: reading, detection and export happen in your browser. You review every area before exporting.
Free for everyone, no account needed. PDF up to 50 MB.
The document
Furnished flat tenancy agreement
Between the undersigned: Landlord: Helen Gardner, 14 Elm Grove, London NW1 6XE Phone: 020 7946 0381, e-mail: helen.gardner@example.com Tenant: Thomas Fletcher, born 4 March 1994 in Leeds Phone: 07700 900461, e-mail: t.fletcher@example.org Property: furnished flat of 42 m², 8 Mill Lane, Bristol BS1 4XQ Monthly rent: £780, bills included, due on the 5th of each month. Rent paid by transfer to account GB82 WEST 1234 5698 7654 32. Deposit: £780, paid with card 4111 1111 1111 1111. Term: twelve months from 1 September 2026, renewable on the same terms. Tenant's notice: one month. Landlord's notice: two months. Signed in London on 28 August 2026, in two copies. Signatures: Helen Gardner, Thomas Fletcher
What leaves your device
Furnished flat tenancy agreement
Between the undersigned: Landlord: Names, Postal addresses Phone: Phone numbers, e-mail: Email addresses Tenant: Names, born Dates of birth in Leeds Phone: Phone numbers, e-mail: Email addresses Property: furnished Postal addresses Monthly rent: £780, bills included, due on the 5th of each month. Rent paid by transfer to account IBANs. Deposit: £780, paid with card Card numbers. Term: twelve months from 1 September 2026, renewable on the same terms. Tenant's notice: one month. Landlord's notice: two months. Signed in London on 28 August 2026, in two copies. Signatures: Names, Names
How it works
Drop the file
A PDF, a Word document or a text file. It’s read inside the tab, scanned pages included, and likely personal data is outlined.
Review every area
Click a box to remove it, add a word to hide every occurrence of it, or draw a box by hand anywhere on a page.
Export the copy
A flattened PDF with nothing under the boxes, or Markdown and plain text with numbered tokens. Saved as document-redacted.
What it finds, and what it misses
Detection does the tedious part. You make the final call: nothing is exported until you’ve reviewed it.
Found automatically
Email addresses.
Phone numbers in international format, or national ones written with separators.
IBANs, checked against their check digits.
Payment card numbers, checked with the Luhn algorithm.
IP addresses.
Postal addresses written in full: number, street, postal code and city.
The names you saved in your list.
Not found automatically: you add them
Other people’s names in running text.
Dates, including dates of birth.
Places and company names.
Case, contract or ID numbers with no recognized format.
Vertical Chinese, Japanese or Korean text.
Formats: PDF, Word and text
What you can drop, and what you get back.
PDF, up to 50 MB
Comes back as a flattened PDF (up to 300 pages), Markdown or plain text. Scanned pages are read by OCR in the browser.
Word, up to 10 MB
DOCX files come back as Markdown or plain text, with every hidden passage replaced by a token or by black blocks.
Text, up to 5 MB
TXT, Markdown, CSV or JSON come back in the same two text formats, with an optional report that counts what was hidden.
Why a black rectangle isn’t enough
Most failed redactions weren’t hacked. Someone drew a black shape over a name in Word or in a PDF editor, sent the file, and the recipient selected the text under the shape and pasted it somewhere else. The shape changes what the page looks like, not what the file contains: the original text, its position and often the document’s author and history are still inside.
Clean Room doesn’t cover the text; it replaces the page it sits on. For a PDF, each page is redrawn as an image at 200 dpi, the boxes are painted into that image, and a new PDF is built from the images alone. There is no text layer, no original object, no title, author or producer field, no attachment and no annotation. Before the download, the tool reads the bytes of the new file and looks for every value you hid; if one were still there, nothing would be downloaded.
The trade-off is deliberate: a flattened page can no longer be searched or copied. That’s the point of a copy meant to be shared. Keep your original, send the redacted one.
Redacting without sending the file anywhere
Free online redaction tools usually upload your document to their servers, process it there and promise to delete it after an hour. For a pay stub, a medical letter, a contract or a court filing, that hour is the problem. Clean Room works the other way around: the page and its code load from our site like any web page, then the file is read, analyzed and rewritten inside your browser tab.
You don’t have to take our word for it. Open the page, switch off your Wi-Fi, and redact a document: it still works, because nothing needs the network once the page is loaded (scanned pages too, once your browser has cached the OCR models). On the page, a counter measured by your browser shows how many requests have left the tab since you dropped the file, apart from the site’s own files and the anonymous statistics beacon. It’s meant to stay at zero.
The tool is free for everyone, guests included. There is no account to create and no quota to count, since our servers never do the work.
The review step is the part that protects you
Automatic detection is good at things with a strict shape: an IBAN has a checksum, a card number passes the Luhn test, an email address has an @. It is poor at things that only make sense to a reader: a first name in the middle of a sentence, the date of an appointment, the name of a street mentioned in passing. No tool can promise to find them all, and we don’t claim to.
That’s why the review is mandatory. Every detection is outlined on the page with its type; one click removes a box that shouldn’t be there. Type a word or a phrase and all its occurrences are hidden at once. Draw a box by hand over a signature, a stamp or a photo. Names you hide often can go into your own list, so the next document starts with them already found.
Scanned pages and languages
A scanned PDF has no text, only pictures of text. Clean Room reads those pages with OCR in the browser, in the language of the interface and in English, up to 60 scanned pages per document. The OCR models are downloaded once from our own site and cached by your browser; the file itself still goes nowhere. A word that OCR misreads can’t be detected, so scanned pages deserve a closer look, and a box drawn by hand covers anything OCR missed.
Email addresses, phone numbers, IBANs, card numbers and IP addresses are found in any language. Postal addresses are recognized in the common European and North American layouts. Vertical Chinese, Japanese or Korean text is not handled yet: draw boxes on those pages.
Frequently asked questions
Is my file uploaded to your servers?
No. The file, its text and what you mark are read and written in your browser tab, and none of it is sent. The page itself loads from our site like any web page, and the anonymous statistics beacon (page views, never any content; it can be turned off in Settings) may still go out. A counter on the page shows any other request sent since you dropped the file.
Is it really free?
Yes, for everyone, without an account. The whole job runs on your device, so it costs us nothing per document and there is no quota. Signed-in users can also open it from any document in their Vault.
Can someone recover the hidden text?
Not from the flattened PDF: each page is an image with the boxes painted in, and the file holds no text layer, no original object and no metadata. In the Markdown and text exports, the hidden passages are replaced by tokens such as [EMAIL_1] or by black blocks. What can leak is what you did not mark, which is why you review every page before exporting.
Does it work with scanned PDFs?
Yes. Scanned pages are read by OCR in the browser, up to 60 per document. A badly printed or blurred word may be misread and then missed by the detection: check those pages closely and draw boxes by hand where needed.
Which languages does it handle?
The interface is available in ten languages. Email addresses, phone numbers, IBANs, card numbers and IP addresses are found in any language; OCR reads the interface language and English. Vertical Chinese, Japanese or Korean text is not handled yet.
Redact your document now
Free for everyone, no account. The file stays on your device.
Open Clean Room