52 private tools · nothing is uploaded

PDF tools, ready on this device

Choose a job. Files stay in this browser — no upload, account or queue.

Three common jobs · one tap away

The tools

52 tools · nothing is uploaded

Nothing appears on this list until it works and can be checked. Every tool has a page that says where it stops.

Pinned and recent tools stay in this browser. This is not an account — there is nothing to sign in to, and none of it is sent anywhere. Your documents are never stored, here or anywhere else.

What a redaction has to mean

Before the document you opened

Account statement page 1 of 2

Account holder
Marianne Okonkwo
Account number
40-21-08 55719246
Address
14 Rowan Court, Bristol BS1 4TR
Closing balance
£2,481.66

Three fields marked: the name, the account number, the address.

After the file this tool hands back

Account statement page 1 of 2

Account holder
removed
Account number
removed
Address
removed
Closing balance
£2,481.66

Searched the finished file for those three phrases: 0 matches.

Not covered — gone: the page is rebuilt as an image, so there is nothing underneath to select. The balance survived because it was not marked. See it run on your own document.

By what you need done

Five groups · each with its own page

Do not take our word for it

Three ways to check

  1. Watch the network

    Open your browser's developer tools, choose the network panel, and use a tool. You will see the tool's own code download from this site, and requests to one other place: analytics.pagefabrica.com, our own page-view counter, which is sent the address of the page you are on and never anything about your document. One thing more, only if you open the Pro door on an OCR tool and start a Deep read: the Deep reader's 660 MB model comes from Hugging Face's servers, once, and that request names the model's files and nothing of yours. Nothing else leaves, and no request carries your file, because no request is made with it.

    This page counts those requests itself, in this tab: it reads the browser's own resource timings, compares each one's origin with this one, and reports how many went elsewhere. Live in this tab: measuring now. The counter counts our own analytics too, and the Deep reader's download. A number that quietly left out the requests this site makes would be worth nothing, so they are in there — the model's download happens in a background thread the page's timings cannot see, so the reader counts its own requests and reports them to the page — and the line says when everything that left went to those declared destinations.

  2. Pull the plug

    Load the page, then switch the device to airplane mode or disconnect the network, and use a tool. It still works: the code it needs is already here, and the file never had anywhere to go. A site that was uploading your document could not do that, and this is the one test that cannot be faked from our side. The exceptions are the readers: an OCR tool's first use downloads its sixteen megabytes, and a first Deep OCR read its 660 MB model, before either works offline — which is why this site no longer shows a "works offline" mark, and says it here instead.

  3. Read what it is built from

    Seventeen open-source libraries do the work. Every one of them is named, versioned and served from this site, with their license texts on the licenses page — and so is the one thing that is not served from here, the Deep reader's model, fetched from Hugging Face behind the Pro door and listed there with its licence. Nothing else runs — there is no certification badge here because we hold no certification, and a borrowed one would be worth less than this list.

Why these tools work this way

The design decisions

Because uploading has nothing to buy

Browsers can already read a PDF, draw its pages and write a new file. Once that is true, sending a contract or a bank statement to a stranger's server is a risk with no benefit attached, so these tools do not offer to take it.

Because a redaction has to prove itself

A black rectangle drawn on top of a name is not a redaction: the word underneath can still be selected and copied. Here every page becomes an image, the file is rebuilt from those images, and then reopened and searched for what you marked. No download appears unless that search returns nothing.

Because the limits matter more than the features

A scanned PDF has no text to convert and this tool says so instead of handing you an empty file. Structure in a converted document is inferred rather than read, and we say that too. The password tool encrypts with AES-128 and says plainly that it is not AES-256, because the encryption is only ever as strong as the password you choose.

Questions people ask first

Do my files get uploaded?

No. Every tool here runs inside your browser tab using code the page has already downloaded. There is no server that receives documents, because the site is a set of static files with no backend at all.

How can I verify that for myself?

Open the browser's network panel and watch while you use a tool: you will see the tool's code load once, and after that only our own page-view counter at analytics.pagefabrica.com, which is sent the page address and nothing from your file — plus, if you open the Pro door on an OCR tool, the Deep reader's model downloading once from Hugging Face. Or load the page, switch your device to airplane mode, and use the tool offline — the counter cannot reach us either, and the tool still works (the first Deep read is the exception: it needs its download first).

Is it really free?

Yes. There is no paid tier today: the Pro door on the OCR tools shows a price to learn whether anyone would pay for the Deep reader, and everything behind it is free during the preview, with no card and no account. No watermark, no sign-up; the fast tools have no page limit beyond what your device can hold, and Deep OCR reads twenty pages at a time. The site is supported by advertising placed away from the tools themselves.

Why is it slow on a big document?

Because your device is doing the work rather than a data center. A long PDF at high resolution is genuinely heavy; the work runs in a background thread so the page keeps responding, and the progress bar shows real pages rather than a spinner.

Which browsers work?

Current versions of Safari, Chrome, Edge and Firefox on phones, tablets and computers. The heavy work needs a modern browser feature for off-screen images; where it is missing, the tools fall back to running in the page and say that they may pause. Deep OCR alone needs Chrome, Edge or Safari 26 on a laptop or desktop.

What does this site remember about me?

Only what you pin, the last few tools you opened, and the settings inside a tool's options panel — all of it in this browser, under one name, and none of it sent anywhere. It is not an account: there is nothing to sign in to and no copy of it exists outside this device. Nothing about a document is ever kept, not its contents and not its name. The button under the tool list clears all of it. Separately, we count page views with our own analytics at analytics.pagefabrica.com. Product measurement is limited to eight property-free event names — work_started, work_completed and the six that describe the Deep OCR preview funnel — the tracker honours Do Not Track, and the privacy page lists them all.