Compress a PDF

Almost always, from pictures. A ten-page report of pure text is a few hundred kilobytes; ten pages photographed with a phone is thirty megabytes. The text is not the problem, and no amount of clever compression will make a text document meaningfully smaller — its size is embedded fonts and page structure, both of which are already compact.

This matters because it tells you what to expect. If your PDF is a scan, an exported presentation, or anything with photographs in it, the savings here are dramatic. If it is a contract typed in Word, it will barely move, and that is not a failure of the tool.

Recompress images — the default

Finds the pictures stored inside the document, re-encodes them smaller, and puts them back. Everything else is left exactly as it was, so text stays selectable and searchable, fonts stay embedded and vector graphics stay sharp at any zoom.

This is what you want in nearly every case. A scanned contract shrinks by 70 to 90 percent and still behaves like a proper document.

Flatten pages — the heavy option

Redraws every page as one flat image and builds a new document from those. It compresses harder and works even on files the first mode cannot touch, but the document stops being a document: no selectable text, no search, nothing a screen reader can read.

Reach for it when a hard size limit leaves no choice, and when the file was already just pictures of pages. Never use it on something the recipient needs to work with.

Image quality is the usual JPEG trade-off. 65 is a good default for scanned text, which tolerates aggressive compression because it is mostly white. Photographs want 75 or above.

Maximum image size caps how many pixels each picture keeps, and it is where most of the saving really comes from. A page photographed at twelve megapixels holds several times more detail than a printer can reproduce, let alone a screen. Capping the longest edge at 1600 pixels is invisible in practice and can shrink a file by more than quality ever will.

Only images already stored as JPEG are replaced. Some PDFs store pictures in other encodings that a browser cannot decode without a dedicated decoder for each one, and guessing at them is how tools corrupt pages. Those images are reported as skipped and passed through untouched.

The tool also refuses to swap in an image that came out larger than the original, which happens routinely with already-optimised files. A compressor that quietly inflates your document is worse than one that does nothing.

Each pass permanently discards detail, and the second pass gains far less than the first while costing just as much quality. If the result is not small enough, go back to the original and lower the settings rather than compressing the compressed file.

Most people compressing a PDF are aiming at a number somebody else chose. The common ones:

• Gmail and Yahoo: 25 MB per message.
• Outlook and iCloud Mail: 20 MB.
• Application and government forms: usually 5 MB, sometimes 2.

There is a trap in those figures that catches people out constantly.An attachment grows by roughly a third on its way through the mail system, because email cannot carry raw binary and encodes it as text. A 24 MB file arrives at a 25 MB limit as about 32 MB, and bounces.

So aim well under the stated number: 18 MB for a 25 MB limit, 14 MB for a 20 MB one. The result panel shows the finished size, so you can check before sending rather than after it fails.

If it will not go small enough, the document is almost certainly a scan. Read the first section again — scans are images, and the recompress mode with a lower quality is the setting that moves them.

The document is opened, rebuilt and handed back entirely inside your browser. That matters here more than for most tools: the PDFs people need to shrink are usually the ones they are about to send to a bank, a landlord, an employer or a government form — exactly the documents that should not be sitting on a stranger’s server.