What this compressor actually does
Be clear about the method before using it, because it is a blunter instrument than the name suggests. Every page is rendered by pdf.js onto a canvas, encoded as a JPEG at the quality your level implies, and that JPEG is embedded into a brand-new PDF at the original page's exact dimensions. The result is a document of pictures.
Preserving the page size in points is why the output still prints correctly and still looks like the original document at a glance. But nothing structural survives: text that was stored as font glyphs is now a photograph of text, so it cannot be selected, searched, copied or read by a screen reader. Hyperlinks, form fields, bookmarks and annotations are gone with it.
That sounds destructive, and for the wrong document it is. For the right document it costs almost nothing. A scanned PDF is already a stack of images — there is no text layer to lose — and this is exactly where the savings are enormous, because scanners routinely embed pages at 300 DPI or more when 150 would be indistinguishable on screen.
What each level does
| Level | Render resolution | JPEG quality | Suited to |
|---|
| Light — best quality | 144 DPI | 82% | Documents that still need to look sharp in print, or pages with fine detail and small type. |
| Balanced — recommended | 108 DPI | 68% | The default. Comfortable on screen, substantially smaller, good for email and uploads. |
| Strong — smallest size | 72 DPI | 55% | Getting under a hard limit. Body text becomes visibly soft; check the preview before using it. |
Two things are moving at once at every level — how many pixels each page is rendered at, and how hard those pixels are then compressed — which is why Strong is much more aggressive than the quality figure alone suggests. The preview of page one after compressing is there precisely so you can judge the result rather than guessing from the percentages.
When this is the wrong tool
Text-based PDFs usually get bigger, not smaller. A document exported from a word processor stores its text as font references and drawing instructions — an extraordinarily compact representation. Replacing that with a JPEG of the same page almost always costs more bytes, and you lose searchable text in the bargain. The tool detects this and tells you the file grew rather than pretending otherwise. If you can select text in your PDF in a normal viewer, this is probably not the tool you want.
Black-and-white scans are a specific trap. A bitonal scan stored with a fax-style or JBIG2 compression is already about as small as a page of text can be, and converting it to a full-colour JPEG both enlarges it and adds grey smudging around the letterforms. If your scan is pure black on white and already modest in size, leave it alone.
There is no way to keep text and shrink images. Desktop software such as Acrobat or Ghostscript can downsample the embedded images inside a PDF while leaving the text layer untouched, which is strictly better when it applies. Doing that requires rewriting the document's internal object structure rather than re-rendering it, and it is not something this browser-based approach attempts. This tool trades that sophistication for never uploading your file.
Other limits. Password-protected PDFs cannot be opened and stop with an error. Every page must be rendered and encoded in turn, so a long document takes real time and real memory. And compression is generational, as always — compress a compressed file and you are re-encoding an already-degraded image.
Getting a file under a limit
Start by asking whether the whole document needs to go. Sending five pages of a ninety-page report, extracted with the PDF splitter, beats any compression setting and does not degrade anything. That is nearly always the better first move when the goal is an email attachment — most mail servers reject somewhere around 20–25 MB, and a trimmed document often clears it untouched.
If the whole document is needed, run Balanced first and look at the preview and the saved percentage. If it is comfortably under the limit, stop — there is no prize for going smaller. If it is close, try Strong and check the preview at full size, paying attention to the smallest text on the page rather than the headings. Keep the original file: the compression is irreversible, and the original is what you will want the next time someone asks for a clean copy.
A useful sanity check before you start: open the PDF and try to select a line of text. If the cursor picks out words, it is a text-based document and this tool will probably enlarge it. If the cursor draws a box over the whole page, it is a scan and you are in exactly the right place. Should you need the words back out of a scan afterwards, OCR can recognise them, and the rest of the set lives on the PDF tools page.