Skip to main content
PDFBase

Redact PDF

Local

Black out text so it is removed from the file, not just covered

All Tools
Tool Details
Input formatsPDF
Output formatPDF
Max file size100 MB
Max files1
PDF render limits

500 pages · 8,192 px/side · 32 MP/page · 100 MP total · 250 MiB output

ProcessingLocal

Privacy First

Your files are processed entirely in your browser. Nothing is uploaded to any server.

How do you redact a PDF?

Drag a box over every word, number or image area that must not be readable, then save. Each boxed page is rendered to an image with your boxes painted into it, and that image replaces the original page, so the covered glyphs are deleted from the file rather than hidden behind a shape. Check the result with the PDF to Text tool on this site before sending it anywhere.

About the Redact PDF

Draw a box over each passage you need gone and save. Every boxed page is rendered to an image with your boxes painted in, and that image replaces the original page, so the covered page content is deleted rather than hidden behind an opaque rectangle. The output is rebuilt without the source document's metadata, attachments or interactive annotations such as comments, form widgets and links. That trade is stated up front: redacted pages become pictures, so their remaining text stops being selectable, while untouched pages keep their text layer. Check the result with PDF to Text, then inspect document properties before sharing — text extraction verifies page text, not every possible PDF data channel.

  • Remove account numbers, national ID digits or medical details from a document before it goes to someone who does not need them, knowing the words are gone from the file rather than waiting under a black box.
  • Prepare a court filing, FOIA release or discovery bundle where a covered-over name or address being copied back out would be a reportable breach.
  • Send a contract to a counterparty with the other side's pricing page blanked out, without pasting pages into an image editor by hand.
  • Audit what you are about to send: the editor counts the words that would survive under a plain covering box, which is exactly how most free redaction tools fail.
  • Work on the sensitive original entirely on your device — the file that most needs to stay private is never uploaded anywhere.

How to use the Redact PDF tool

  1. 1

    Upload PDF file

    Select PDF file from your device or drag file into the upload area.

  2. 2

    Draw a box over each passage to remove

    Drag a rectangle over every word or number that must not survive. Each box rasterizes its page on save, so the glyphs underneath are deleted from the file rather than hidden.

  3. 3

    Save redacted PDF

    Save when every region is covered. The note names the pages that were rasterized, because those pages become images and their remaining text stops being selectable.

  4. 4

    Verify the removal

    Open the saved file in the PDF to Text tool and search for the removed words. No match confirms they are absent from extractable page text. Also inspect document properties and attachments: text extraction alone cannot certify every data channel in a PDF.

Why do most blacked-out PDFs still contain the text?

Because a filled rectangle is a shape drawn on top of the page, not an edit to it. The glyphs underneath stay in the content stream exactly as produced, so selecting the area and copying, or running any text extractor, reads them straight back. Documents released by courts and government departments have failed this way repeatedly; the words were never removed, only painted over.

What does true redaction change about the rest of the page?

A redacted page no longer has a text layer at all: everything on it becomes part of one image at 200 DPI, so surviving sentences stop being selectable or searchable and the file grows because a bitmap is larger than drawing instructions. This is per page — a document with two sensitive lines in a hundred-page file rasterizes exactly those pages and leaves the other ninety-eight untouched.

How do you check that a redaction actually worked?

Use more than one check. First run the saved file through a text extractor and search for the terms you removed; on this site, open the redacted copy in PDF to Text. A match proves the page redaction failed, while no match confirms the term is absent from extractable page text. Then inspect document properties, comments, form fields and attachments separately, because a text extractor does not read every PDF data channel. PDFBase rebuilds its redacted output without source metadata, attachments or interactive annotations, but the layered check is still the safer habit for files produced anywhere.

Does redacting work on scanned documents?

Yes, with one caution. Scanned pages are already images, and the box is burned into the rendered pixels, so whatever it covers stops being visible in the output regardless of whether there was ever a text layer underneath. The caution is coverage: a rectangle that clips half a line of handwriting leaves half a line of handwriting, so size boxes generously and zoom in before saving.

Are my files uploaded when redacting a PDF?

No. Redaction is performed by code running in this tab: pdf.js renders each marked page, your boxes are composited onto that render in memory, and pdf-lib writes the replacement pages out again. The most incriminating version of the document therefore never touches a network connection — which matters more here than anywhere else, because a redaction uploaded for processing has already defeated its own purpose.

How to do this offline, on the command line

True redaction means deleting glyphs from the content stream. Painting black rectangles with an ordinary PDF library reproduces exactly the failure this tool exists to prevent.

PyMuPDF

Python: pip install pymupdf

python -c "import fitz; d = fitz.open('statement.pdf'); p = d[0]; p.add_redact_annot(fitz.Rect(72, 96, 320, 112)); p.apply_redactions(); d.save('statement-redacted.pdf')"

apply_redactions strips the text inside each rectangle out of the page before saving — the same destructive guarantee this page gives. Coordinates are points measured from the top-left.

pdftotext (verification)

macOS: brew install poppler · Debian/Ubuntu: apt install poppler-utils

pdftotext statement-redacted.pdf - | grep -c '4929'

Zero matches confirms the number is absent from extractable page text. It does not inspect metadata, comments, form values or attachments; check those separately before sharing a sensitive file.

Frequently Asked Questions

Is the blacked-out text recoverable from the saved file?
The original page content under each box is not kept. Each redacted page is replaced by an image of itself with your boxes painted in, and the original text objects are dropped from that page. The output is rebuilt without source metadata, attachments or interactive annotations such as comments, form widgets and links. Verify the page text in PDF to Text, then inspect document properties as a separate safety check.
Why did my redacted PDF get larger?
Pages carrying a box are now stored as 200 DPI images instead of compact drawing instructions, which costs space. Pages without boxes are kept exactly as they were, so only the pages you actually redacted grow.
Can I redact a scanned PDF?
Yes. The box is burned into the rendered pixels of the page, so whatever it covers disappears even though a scan has no text layer. Draw each box generously: one that clips an edge can leave part of a word visible.
Can I undo a redaction after saving?
No, and that is the point — the original content under a redaction box is not stored in the output page. Your original file is never modified or uploaded, so keep it and re-redact from that copy if you need a different version.
Does redacting need an internet connection?
Only to load this page. Rendering, boxing and writing the new file all happen inside your browser, so once the page is open you can disconnect and continue working offline.

Need More Tools?

Explore our collection of 51 free PDF tools, all with privacy-first processing.

Browse All Tools