Compress PDF

How the Compress PDF works

Most oversized PDFs are big for one reason: photos. A scanned page, a phone picture of a receipt or a brochure exported at print quality can hold images of 4000 pixels or more across, saved at near lossless quality. On screen, and on most office printers, nobody can see the difference between that and an image a third of the size. The rest of a PDF, the text, fonts and vector drawings, is usually small already. So a good compressor spends its effort on the pictures and leaves everything else alone.

How to compress PDF files here

The tool opens the file, finds every embedded image and checks whether it is a photo it can safely re-encode: 8-bit greyscale or colour, stored as JPEG or as compressed raw pixels, with no palette, stencil or colour-key mask. Each of those photos is scaled down so its longer side fits the level you chose, then saved again as a JPEG at that level's quality. If the new version is not smaller than the old one, the original image is kept. Soft transparency masks are carried across unchanged, so logos with transparent edges still look right.

It then looks for data streams that were stored with no compression at all, which some older scanners and export tools still produce, and packs them with the same Deflate method used in ZIP files. That step is lossless. Finally the file is rewritten with object streams, a PDF 1.5 feature that groups the small structural objects together and compresses them, which typically trims a few more percent.

Choosing a level

• Balanced suits most documents. Photos up to 1600 pixels on the long side at quality 70 stay sharp on a laptop screen and print cleanly at A4 or Letter size.
• Strong is for hitting an upload limit, such as a job portal or council form that caps files at 2 MB. Photos up to 1100 pixels at quality 50 can look soft if zoomed in, but text in a scan stays readable.
• Light keeps more detail for documents that will be printed large or zoomed into, like a property brochure or a portfolio. The saving is smaller.

If a file is already well compressed, nothing useful can be removed and you get the original back unchanged, with a note saying so. That is better than a "compressed" copy that is somehow bigger.

What the numbers mean

The result shows the original and new size and the percentage saved. "Photos re-encoded" reads as "x of y": y is how many images qualified, x is how many came out smaller and were swapped in. "Streams packed" counts the uncompressed data blocks that were deflated. A scan made of one big photo per page tends to shrink by 60 to 90 percent. A text report with a few charts may only lose 5 to 20 percent, because there is little to squeeze.

Metadata

By default the tool also drops the document information dictionary and the XMP metadata block. These hold the title, author, software name and edit dates. They are rarely large, but removing them saves a little space and avoids sharing details you may not have meant to send. Choose Keep if the title or author must stay, for example on a published report.

What it does not change

• Text and fonts. Text stays as real, selectable and searchable text. Fonts are not subset or swapped, so the layout cannot shift.
• Line art, charts and diagrams. Vector graphics are already compact and are left alone.
• Black and white scans stored as fax (CCITT) or JBIG2 images. These formats are already far smaller than JPEG would be for text, so converting them would make the file bigger.
• Forms, links and bookmarks. They are carried over untouched.

Tips and pitfalls

• Compress once, from the best copy you have. Running a JPEG through compression again and again loses a little more detail each time.
• If a scanned page looks blotchy after Strong, try Balanced. Text with coloured highlights shows JPEG artefacts first.
• Encrypted or password-protected PDFs cannot be rewritten until the protection is removed.
• If the file is still too big, split it into parts or delete pages that are not needed, then compress what is left.
• A PDF that is mostly one huge embedded font or attachment will not get much smaller here, because neither is touched.

Privacy

Compression happens in the browser tab using the open-source pdf-lib library and the canvas JPEG encoder built into your browser. The file is not uploaded, stored or logged anywhere, which matters for payslips, medical letters, contracts and ID scans that people most often need to shrink.

Related tools