Compress PDF
Shrink a PDF by recompressing the photos and scans inside it: the text stays text, selectable and searchable. Everything happens in your browser.
One PDF file
How it works
There are two methods. Images only, the suggested one, leaves the text as it is (you can select it, copy it, search it) and recompresses the photos and scans inside the PDF: each one is reduced to the resolution it needs for the size it is shown at on the page, so a big photo printed small loses the pixels nobody sees. With Light the images stay at 220 dpi, with Medium at 150 and with Strong at 110. Diagrams, logos and images of a kind that is not touched here (CMYK for print, JPEG 2000, those inside comments) stay as they are.
Every page as an image instead redraws each page as a lower-quality photo: it lightens documents made only of scans more, but the text becomes a drawing. Start with Medium and move to Strong only if you need a much lighter file.
An important note
With Every page as an image, in the compressed PDF the text will no longer be selectable or searchable, which is why the suggested method is the other one. Before offering you the file, the page opens it again and checks that it has the same pages and exactly the same text as the original, and it never offers it if it weighs more than the one you have. If the PDF is digitally signed, the signature is no longer valid in the compressed file. If the file is still over the portal's limit even after compressing, you can break it up with Split a PDF by size.
Where the size is, almost always
In a heavy PDF ninety per cent of the space is in the images, not the text: a thousand pages of plain text fit in a few megabytes, a single colour scan at 600 dpi takes five. Compressing a PDF therefore means, nearly always, recompressing its images.
The other heavy item is embedded fonts: if every part of a composite document brings its own, you reach several megabytes of typefaces alone. It is why a file born from merging twenty documents weighs more than the sum of its pages.
The number that counts is dpi, not quality
Lowering JPEG «quality» gains less than people think; the real saving comes from reducing resolution. A scan at 600 dpi holds four times the pixels of the same page at 300, and 300 dpi is already more than enough to print a text document.
There are three practical thresholds: 300 dpi for printing, 150 dpi for reading on screen or emailing, 72 dpi for a preview only. Going below 150 on a document with scanned text makes it tiring to read and useless for character recognition.
When compressing is pointless
If the PDF is made of real text and not scans, compressing gains almost nothing, because text is already compressed and takes very little. A 300 kB file of pure text pages is already at its minimum.
And if the document is to be archived or sent to a printing house, compressing is exactly what not to do: compression is irreversible, and a reduced image cannot be brought back. In that case you keep the original and compress only the copy you send.
Nearby tools
If the limit to respect is a precise one, Split a PDF by size makes parts already under the threshold. For Office files the equivalent is Compress an Office file, and for images alone Compress image.