When this works well, and when it does not
PDF compression is really image compression in disguise. Most oversized PDFs are large because they contain scanned pages stored as high-resolution photographs. Reducing the resolution and re-encoding those pictures is where nearly all the saving comes from.
- Works very well: scanned documents, photographed forms, PDFs exported from a phone scanner app, presentations with large photos.
- Works poorly: PDFs that are mostly text created directly from Word or LaTeX. Those are already small, and rebuilding them as images can make them larger while losing searchable text.
How the target size mode works
Rather than making you guess at DPI and quality settings, target mode renders your pages once and then searches for the highest quality that still lands under your limit. If quality alone cannot get there, it steps the resolution down and tries again.
This means it takes longer than picking settings by hand, particularly on long documents, but it removes the usual cycle of compressing, checking the size, and compressing again.
Choosing a resolution manually
| Setting | Best for |
|---|---|
| 100 DPI | Reading on screen, email attachments |
| 150 DPI | Most upload forms, general use |
| 200 DPI | Documents that may be printed |
Greyscale is worth trying on any black-and-white text scan. Colour information on a page that has no colour costs bytes for nothing, and switching it off often cuts the file by a third or more.
Questions
Will my text still be searchable afterwards?
No. Because each page is rebuilt as an image, text selection, copying and search stop working. If you need searchable text, keep the original and send that instead.
Is my document uploaded?
No. The whole process runs in your browser. This matters more for PDFs than for most files, since they often contain identity documents, medical records and financial statements.
Why did my PDF get bigger?
Your original was probably a text-based PDF rather than a scan. Those are already efficiently stored, and converting them to images adds bytes. Keep the original.