PDF Toolbench: merge, redact, sign, OCR & compress PDFs in your browser
PDF Toolbench
A PDF utility that runs entirely in your browser. Merge files, reorder, rotate, and delete pages, pull out a range, add text or a signature, fill in forms, redact a page so the hidden text is actually gone, make a scanned PDF searchable with OCR, copy the text out cleaned up for pasting into an AI chat, add watermarks or page numbers, export pages as images, strip metadata, shrink the file, or lock it with a password. Everything runs locally on your device, your PDFs are never uploaded, and Ctrl+Z undoes mistakes.
🔒 Everything runs locally in your browser. Your PDF is never uploaded and never leaves your device. (I use Google Analytics to see whether anyone visits this page, but it cannot see your files or anything you do with them.)
A caveat: this is a personal side project and I have not tested it exhaustively. I make no guarantees; keep your originals and check the output before relying on it.
Double-click any page to add text, sign, black out, or fill forms.
or click to choose
Loading the PDF engine…
A single click selects a page; double-click it (double-tap on a phone), or click its ✎, to open it for editing. Drag a thumbnail to reorder, pressing and holding first on a phone. Ctrl+Z, or the Undo button, reverses the last change.
Built on open-source libraries: pdf.js by Mozilla (Apache 2.0) for rendering, pdf-lib (MIT) for editing, tesseract.js (Apache 2.0) for OCR, and qpdf (Apache 2.0, compiled to WebAssembly) for encryption. All are served from this site, never from a third-party CDN. Thank you to their authors.
Common questions
How do I redact a PDF without uploading it?
Add your PDF above, double-click the page (or click its ✎ button), choose the Redact tool, and draw boxes over anything you want removed. The boxes apply as soon as you draw them and stay with the page while you work (move them, delete them, or undo with Ctrl+Z). Real redaction cannot just draw a black rectangle, because the text underneath would still be in the file, so when you download every page with boxes is re-rendered as an image with the boxes painted on, replacing the original page. The hidden text is actually gone. You can verify it yourself: try to select or search the redacted area in the downloaded file. Nothing is uploaded at any point.
What do the download buttons and the Pages box do?
Download all saves the whole document as one PDF: every page shown in the grid, in that order, with all edits applied. Download selected saves only the selected pages (or the range typed in the Pages box) as their own PDF, and leaves the document here as it is. Split saves each page as its own PDF inside a .zip. Shift-click selects a run of pages, and the Pages box takes a range like 1-5,9,12- for both the selected-pages download and Split. Adding several files merges them into one document, in the order shown, and images are turned into pages.
How do I make a scanned PDF searchable?
Add the scan, click "Make searchable (OCR)", and download. An OCR engine running in your browser reads each page and adds an invisible text layer on top, so the file looks identical but you can select, copy, and search the text. It reads printed English and works best on clean scans. Handwriting usually does not work.
What does "Copy clean text" do?
It pulls the text out of the document, cleaned up for pasting into an AI chat or anywhere else. Running headers, footers, and page numbers are removed by a heuristic: a line that repeats in about the same spot on most pages, or a line that is just a page number, is dropped. Words broken across a line wrap are rejoined, and the rest is reflowed into paragraphs. It is a heuristic, not a layout engine, so complex layouts like tables and multi-column pages can come out in the wrong order, right-to-left scripts like Arabic and Hebrew come out with the words reversed, and scanned pages need "Make searchable" first since there is no text to read yet. Text you added with the Text tool is included; values you typed into form fields are not. Nothing is uploaded; it reads the text already open in your browser.
What happens to my files?
They are opened and edited in your browser's memory and never sent anywhere. The processing libraries are served from this site, and your document stays on your device. Close the tab and everything is gone. Saved signatures, if you use that feature, live only in your browser's local storage.
Why is this free?
I wanted a PDF tool I could trust with sensitive documents, and the only honest way to build that is to keep the files on your device. Since your browser does all the work, hosting this page costs me almost nothing. There is no account, no upsell, and no server that could see your files even if I wanted it to.
Each box is kept and shown on the page as soon as you draw it. Draw as many as you like, move them, delete them, or press Ctrl+Z. The page itself keeps its full quality while you work, and stays untouched if you remove every box.
The content is removed when you download. Drawing a black box on top is not enough on its own: the text underneath would still sit in the file and could be copied straight back out. So every page that has boxes is re-rendered as an image with the boxes painted on, and that image replaces the page. The covered content is then actually gone, and because the page is now an image, its text is no longer selectable or searchable.
This shrinks the file by rendering every page to an image at the chosen quality and rebuilding the PDF from those images. It's most effective on scans and image-heavy documents. On text-only or already-optimized PDFs the result can come out larger, not smaller, and the status line reports the size change either way.
Trade-off: like redaction, the pages become images, so text is no longer selectable or searchable. (Your original file is untouched; this downloads a copy.)
The password is used only on your device, to open the file.
The watermark is drawn into the PDF on download. Rotated pages are flattened upright first. Ctrl+Z removes it again.
Numbers follow the page order shown in the grid, update automatically if you reorder, and are drawn into the PDF on download. Rotated pages are flattened upright first. Ctrl+Z removes it again.
Renders pages to image files. One page downloads as a single image; several pages are bundled in a .zip.
This reads the text on your pages with an OCR engine that runs in your browser (tesseract.js) and adds an invisible text layer on top. The pages look exactly the same, but the text becomes selectable, copyable, and searchable. Download afterwards to save the result.
It reads printed English and works best on clean, straight scans. Handwriting usually does not work. The first use downloads about 6 MB of OCR engine code from this site (no third-party servers); your browser keeps a copy for next time. Your pages are processed locally and never uploaded.
Extracts the text on , with running headers, footers, and page numbers heuristically removed, line-wrap hyphens repaired, and paragraphs reflowed into plain text, ready to paste into an AI chat or anywhere else. This reads the text layer already open in your browser; nothing is uploaded.
Builds the document and encrypts it with AES-256 using qpdf running in your browser, then downloads it. Opening the file will require the password. The first use downloads about 1.3 MB of engine code from this site; your PDF never leaves your device.
Careful: there is no way to recover the password later. Write it down.