A demonstration, on your own document
Drawing a black box over a name, a grade or an amount does not remove the text from the file: the box is a drawing placed on top, and the text goes on living underneath, selectable and copyable. Drop a document you believe is redacted — VellumPDF shows you what can still be read under it. Nothing is sent anywhere, which is exactly what lets you try it with a real document rather than a sample.
Local analysis — no file sent to any server
Drop a PDF you believe is redacted
The analysis runs in your browser: the file goes nowhere
A PDF is not an image: it is a list of drawing instructions in which text keeps its characters and its coordinates. A box added on top is one more instruction, executed after the others. On screen it covers; in the file it erases nothing. A copy and paste, a text extraction or a plain search brings the covered word back in a second. That is how patient names, salaries and addresses resurface from reports published as redacted.
We blacked out a fake Swiss social security (AVS) number in a made-up certificate in five ways, then checked whether it could still be read. Black highlighter in Word, then export to PDF: the number can still be copied. Black rectangle drawn in Word: same result. Rectangle filled with black in Preview on a Mac: the number can still be copied, even though Preview warns that the content underneath is not removed. Comment rectangle with a black fill in the free version of Acrobat: the number can still be copied, with no warning. Only Preview’s Redact tool removed the text. Four gestures out of five fail, and Word’s highlighter, the most common of them, warns about nothing.
The text layer has to go, not be covered. VellumPDF’s redaction renders every page that carries an area as an image, paints the chosen areas onto it, then replaces the page with that image: the targeted text no longer exists in the file produced, because the redacted page has no text left at all. The trade-off is deliberate and worth knowing: a redacted page is neither selectable nor searchable; the other pages stay intact, or become images too for those who prefer the safest option. That is the price of a real blackout.
The analysis picks up dark rectangles drawn in the page and square annotations, then checks on the rendered image that the area really is opaque before announcing anything: a dark title band carrying light text is not a hiding place, and it is not counted as one. If your document is a scan with no text layer there is nothing to extract, and the page says so rather than reassuring you wrongly. The first forty pages are examined.
No. The analysis is computed by your browser, on your device, like every other VellumPDF tool. You can cut your internet connection before dropping the document: the demonstration still works, and that is the simplest way to check the promise.
Do not pass it on, and do not work from the version with the box: go back to the original document and run it through the redaction tool, which genuinely removes the text. If the document has already circulated, treat the information as read — a shared file cannot be recalled.
Two different cases hide behind that result, and the page tells them apart. Either the masking was done properly and no text remains under the areas. Or the document holds no text at all — an unrecognised scan, for instance — in which case the analysis has nothing to read and proves nothing.
No file is uploaded, no account, no advertising, no daily limit: nobody pays with their data, and there is nothing to sell behind it. The site is written and maintained in Carouge, Switzerland, by the people who built it; the domain name, the certificates, other tools and the hosting are paid for every year. A donation, whatever each person can afford, helps pay for that upkeep and keeps every tool open to everyone.
A one-off amount of your choosing, through Stripe. Nothing is unlocked in return: it is already free, for everyone.
A black box drawn over a name, a grade or an amount does not remove the text from the file: it is a drawing added on top, executed after the text, which stays selectable and copyable underneath. This page demonstrates it on your own document: it picks up the dark areas, checks on the rendered image that they really are opaque, then shows the text still hiding under them. The analysis is computed by the browser, with nothing uploaded, and the result leads to the redaction that genuinely removes the text layer.