Skip to content
Server-sideDeleted in 30 minutes

Safely Redact One PDF Page

Redact a rectangle on one PDF page without leaving that page's old text underneath a black drawing overlay. The browser renders the chosen page with pdf.js, paints an opaque black rectangle into the page pixels, and rebuilds the PDF with that page replaced by a raster image. Other pages are copied unchanged and remain text-based, so this is a page-only redaction utility rather than a claim of complete document-wide redaction.

Use this without the search next time. Prathom Workbench puts Prathom's tools in your toolbar.

Add to Chrome — free

PDF Redactor

Mark one or more areas on any page. Marked pages are rebuilt as flat images — the original text, links and metadata underneath are discarded, not just covered.

What it does

  • Opaque pixels replace the selected page's old text layer
  • Reopens the output to verify the selected page has no text layer or annotations
  • Coordinate form uses PDF points from the lower-left origin
  • Visible preview guide shows the proposed rectangle
  • Other pages are copied without rasterizing them
  • Processing stays in the browser and the source is unchanged

How to use PDF Redactor

  1. 1

    Open the PDF

    Drop a PDF into the tool. The browser reads the page count locally and renders the first selected page for inspection.

  2. 2

    Enter the page coordinates

    Choose one page number, then enter X, Y, width, and height in PDF points. X and Y use the lower-left corner as the origin, and the rectangle must fit inside the page.

  3. 3

    Check the preview

    Compare the outlined guide with the visible content before creating the file. Change the numbers or page and wait for the preview if the page orientation or target area is not right.

  4. 4

    Create and inspect the copy

    Press Redact selected page, download the new PDF, and open it in a reader. Keep the source until you have checked every sensitive area and confirmed whether more pages need separate work.

How it works

The browser uses pdf.js for the part pdf-lib cannot do: page rendering. After the PDF is selected, the tool opens it in a worker, obtains the chosen page's viewport, and paints a white-backed version onto a canvas. The form values stay in PDF-point space, with the origin at the lower-left of the preview. The exported pdfRectToCanvasRect helper flips the vertical direction and multiplies by the render scale, so the same rectangle lands on the correct pixels at preview and output resolutions.

The preview shows the proposed rectangle as an outline with a light guide fill, and moving it changes nothing in the document. When the action button is pressed the page is rendered again at 144 DPI and an opaque #000000 fill is painted into the calculated pixel rectangle. PNG is used rather than a lossy format so the black boundary and surrounding document pixels do not gain compression artifacts.

The rebuild is deliberately asymmetric. A new pdf-lib document receives a PNG-backed replacement page at the rendered page dimensions. The source page is never copied, which means its old text streams, form widgets, and page annotations are not present on that replacement. All non-selected pages are passed through copyPages and added in their original order. That keeps their selectable text and normal PDF behavior, but it also means those pages have not been redacted by this operation.

Security limits

This is safer than a black rectangle overlay for the selected page, not a complete document sanitizer. It does not search for names, OCR text, repeated identifiers, document metadata, embedded files, comments, bookmarks, JavaScript, or matching content on other pages. A recipient may still learn sensitive information from an unredacted page, a duplicate in an attachment, or the original file retained elsewhere. The rasterized page also loses its text layer, links, accessibility tags, and vector detail. Inspect the result in more than one PDF reader, search for the sensitive value in the output, and keep the source private. For legal, regulated, or adversarial disclosure work, use a specialist redaction process that audits the entire document and its file structure.

When to use it

Use this tool when one visible region on one page needs to be covered before a quick handoff, such as an account number on an invoice, a signature on a scanned application, or a handwritten note on a form. It is also useful for a small preview copy where the untouched pages should remain searchable. Choose a document-wide redaction product instead when a person, identifier, or topic can appear on several pages, in hidden layers, or in attachments.

Examples

Remove an account number from an invoice

invoice-march.pdf — page 1 of 2 X: 356 · Y: 676 · width: 132 · height: 24 points
invoice-march-redacted-page-1.pdf Page 1 is a raster image with an opaque black rectangle; page 2 is copied as PDF content.

The invoice's account number is covered in the pixels that recipients see, while the terms and payment instructions on page 2 remain selectable. That split is useful when only a single field needs removal and searchable text on the rest of the invoice is still wanted.

Cover a signature on a scanned application

application.pdf — page 3 of 7 X: 72 · Y: 88 · width: 210 · height: 58 points
application-redacted-page-3.pdf Page 3 is rebuilt from a 144 DPI raster with the signature area painted black.

A scanned signature is image content, so a PDF text overlay would not remove it at all. Rendering the complete page first means the black paint covers the scan, its OCR layer if present, and the visible signature together on that selected page.

Frequently asked questions

Does a black rectangle really remove the text underneath?

In this tool, the selected page is not given a PDF drawing rectangle on top of its old content. pdf.js renders the page to an opaque canvas, the black rectangle is painted into those pixels, and a fresh PDF page containing only the PNG is added. The selected page's former text objects and annotations are therefore not copied into the replacement page. Always inspect the downloaded file before distribution.

What does the verification message check?

After creating the file, the browser reopens the output and checks the selected page for extractable text and PDF annotations. A pass means those removable objects are absent from that page; it cannot OCR the raster pixels, detect a sensitive image, or search other pages and attachments.

Are all pages in the output safely redacted?

No. Only the one page selected in the form is rasterized and replaced. Every other page is copied as a normal PDF page and remains text-based, with its original searchable content and possible hidden objects. This tool does not find matching words elsewhere, inspect metadata, remove attachments, or promise document-wide redaction; repeat the workflow for additional pages and review the whole document.

What do X, Y, width, and height mean here?

They are PDF points in the orientation shown by the preview, not browser pixels. X measures from the left edge, Y measures from the bottom edge, and width and height expand the rectangle rightward and upward. There are 72 points in an inch. Use the outline as a visual check, because a coordinate that is mathematically valid can still cover the wrong nearby words.

Does rasterizing reduce the quality or accessibility of the page?

Yes, and that is an intentional tradeoff for removing the selected page's recoverable text layer. The page is rendered at 144 DPI and becomes an image, so text selection, search, links, accessibility structure, and vector sharpness on that page are not preserved. Other pages keep their normal PDF content. For high-resolution printing or accessible archival copies, use a specialist redaction workflow and validate its output.

Is my PDF uploaded or overwritten?

No. pdf. The marked pages are rasterized on the server so their text objects cannot survive, and your original file is never modified. The result is a new download assembled in memory. Local processing does not remove the need for document review: browser memory, screenshots, copies of the source, other pages, metadata, and attachments are outside the narrow page replacement performed here.