Compress PDF files online

Almost certainly because it is a text document rather than a scan. Text, fonts, and vector graphics inside a PDF are already compressed when the file is created, so there is little left to squeeze. The only way to shrink such a file substantially is to convert the pages into images, which is what the Strong setting does — at the cost of selectable text.

Balanced rewrites the file structure and re-encodes image data where it can. On a document that's mostly text, that leaves it selectable and searchable and the file barely changes size. On a scanned document — mostly image data to begin with — Balanced rasterizes every page the same way Strong does, just at a higher quality and larger size. Strong always renders every page to a JPEG and rebuilds the PDF from those images, for the smallest result regardless of what the file is made of: no selectable text, no working links, and slightly softer edges when zoomed.

No. Compression runs in your browser using JavaScript and WebAssembly, so the file stays on your device. You can confirm this by opening your browser devtools Network tab while compressing — there are no upload requests.

On Balanced, a text document is unchanged, since nothing in it gets re-encoded. A scanned document is rasterized to JPEG at a fairly high quality, so expect a little softening. On Strong, every page becomes a smaller, lower-quality image, so the drop is more noticeable everywhere. If Strong looks worse than you want, run the original again on Balanced.

We give you your original back instead. Some PDFs are already optimally packed, and re-saving them can add a few bytes. Handing you a bigger file and calling it compression would be dishonest, so we do not.

Not directly. A PDF with an open password has to be unlocked first, using a password you know. Once unlocked, it compresses like any other file.

Usually because they run on servers with tools like Ghostscript, which can rebuild content streams and subset fonts in ways a browser cannot. That genuinely does better on text-heavy PDFs. The trade is that your document has to be uploaded to a stranger. For scanned files — the ones people most often need to shrink — the gap is small.

Strong, and Balanced on a mostly-scanned document, both work by rendering every page as an image before rebuilding the file — and a long, high-resolution scan can render to far more image data in memory than the original PDF's own file size suggests. We check for that before starting rather than partway through, and say so rather than letting the tab crash. Splitting the document into smaller parts first works around it.