pdfpix

Browser vs server PDF tools

A browser tool reads your file in the tab and never sends it anywhere. A server tool uploads it, does the work on someone else's machine, and sends a result back. Both are legitimate. What changes is who holds the file, and which jobs are possible at all.

What actually differs

In a browser, the file is opened by code already downloaded to your machine. There is no upload, no queue, no copy on a disk you cannot see, and no retention policy to read, because there is nothing to retain. Work offline and it still runs.

On a server, your file travels, sits somewhere while it is processed, and is deleted on a schedule the vendor sets. Xodo states files are permanently deleted after one hour where upload is required, and TinyWow states the same hour for processed and unprocessed files, both checked 20 August 2026.

The jobs that genuinely need a server

Text recognition on scans. The models are large, the work is heavy, and doing it in a tab is either impossible or unbearably slow. pdfpix has no text recognition for exactly this reason. If your PDF is a picture of words, you need a server or a desktop application.

Office conversion is the second: turning a PDF into Word or Excel needs layout engines nobody ships to a browser. Long batches are the third, because a tab that grinds for twenty minutes will be closed by its owner. Signing with a certificate is the fourth, since the trust anchor has to live somewhere you do not control.

What a browser does better than people expect

Compression to a size target runs four methods and verifies every candidate, locally. Merging, page organising, rotation and cropping, image assembly, page export, stamping, redaction and metadata removal all work the same way.

Median compression time on the public test corpus was 2512 milliseconds, with the ninetieth percentile at 3968. That is not a compromise made for privacy. For these operations the tab is simply fast enough, and the round trip to a server would cost more than the work.

Self hosting is the third option

You can run a server that is yours. Stirling PDF publishes more than sixty tools under the line your PDFs, your servers, and states files never leave when run on your own infrastructure or air gapped, with more than thirty million downloads.

PDFsam Basic is the desktop equivalent: free, open source under the AGPL, on Windows, macOS and Linux since 2006, stating no watermarks, no size limits and that files stay on your computer. Both need a machine and some setup, which is the real cost.

How to tell which one you are using

Open the network panel in your browser's developer tools before you pick a file, then run the tool. If your document leaves, you will see a request carrying it. Nothing else a page says about privacy is as reliable as that check.

A second signal is behaviour offline. Disconnect and try the tool: a browser tool keeps working, a server tool cannot. How it works describes what runs where on this site, and the privacy page states what analytics records, which is route names and stage outcomes, never document bytes, file names or text.

Questions people ask

Which PDF jobs genuinely need a server?
Text recognition on scans, office format conversion, very long batch runs and certificate signing. Everything else on the tool list, including compression to a size target, runs in a tab at usable speed.
How do I know if a PDF tool is running in my browser?
Open the network panel in developer tools before choosing a file and watch for a request carrying the document. Or disconnect from the internet: a browser tool keeps working, a server tool stops.
Is a browser PDF tool slower than a server one?
Not for these operations. Median compression time on the public test corpus was 2512 milliseconds, ninetieth percentile 3968. A server has to receive the upload before it starts, which usually costs more than the processing.
Are server PDF tools less private by definition?
They hold your file for some period, so you are trusting a policy rather than a fact. That is reasonable for ordinary documents and a poor fit for medical, legal or financial ones. Self hosting removes the third party without giving up server capability.

The tools this compares