Find Out What Is Actually Large
Before compressing anything, it is worth knowing what is taking up the space, because the answer determines whether compression will help at all.
| What is in the file | Typical share of size | Does compression help? |
|---|---|---|
| Scanned pages (images) | Almost all of it | Enormously, this is the case compression is for. |
| Photographs or screenshots | Most of it | Substantially. |
| Embedded fonts | A few hundred KB | No, they are replaced by the page image. |
| Text and vector graphics | Very little | No, the file comes back larger. |
The rule of thumb follows directly. A scanned document is often 95% image data, and redrawing it at a lower resolution can remove three quarters of the file. A text-only PDF exported from a word processor is already close to minimal, and redrawing its pages as images makes it several times larger instead.
So if a 200 KB text document comes back bigger, nothing is broken. There was nothing left to remove, and the only way to find that out was to redraw the pages. Keep the original.
The Four Settings Are Resolution Targets
Each quality setting renders every page to an image at a target resolution and stores it as a JPEG. The file you get back is that stack of images and nothing else. That is the whole mechanism, and choosing between the settings is really choosing what the document is for.
- Maximum compression, 72 DPI. Screen resolution. Smallest possible file, appropriate for email attachments and anything that will only ever be read on a display. Printed, it will look visibly soft.
- Balanced, 150 DPI. The sensible default. Still crisp on screen, acceptable on paper, and typically the largest reduction you can make without anyone noticing.
- High quality, 300 DPI. The standard for printing. Choose this when the document will be printed and read closely rather than filed.
- Minimal, 300 DPI at the highest image quality. The same resolution as the previous setting with far less JPEG smoothing, for documents where fine detail matters more than file size.
The distinction between the last two is worth understanding: they render at the same resolution and differ only in how hard the JPEG encoder is pushed. The fourth keeps detail that the third smooths away, in a noticeably larger file.
Compression Is Not Reversible
Downsampling discards pixels permanently. A 300 DPI scan reduced to 72 DPI cannot be restored, running the result back through at a higher setting produces the same soft image in a larger file.
So keep the original. Compress a copy, look at the result at full zoom before you send it, and pay particular attention to small text in scanned documents, which is where downsampling shows first. Text that was marginal at 300 DPI becomes unreadable at 72.
The other thing to know is that the document stops being searchable. A scanned PDF that has been through optical character recognition carries a text layer behind the image, and because every page here is redrawn from scratch, that layer goes, along with any form fields, links and annotations. What comes out is a stack of images. If the document needs to stay searchable or fillable, this is not the tool for it.
When to Reach for Something Else
Compression is not always the right tool. If the file is large because it contains fifty pages and you need three, splitting it is a better answer than degrading all fifty. If it is large because a single high-resolution photograph was embedded at full size, replacing that one image is more precise than downsampling everything.
And if the goal is emailing something and the attachment limit is the real constraint, a shared link avoids the problem entirely without touching the document.
This tool runs in your browser using a Web Worker, so the file never leaves your machine and processing does not freeze the page. That matters for contracts, medical records and anything else you would not upload to a stranger's server, which is most of what people compress.