What this tool actually does
It re-renders each page at a lower resolution, encodes it as a JPEG, and assembles a new PDF from those images.
That is worth understanding before you use it, because it explains both why it works so well on one kind of file and why it is the wrong tool for another.
Best case: a scan
Scanned PDFs are the files that are actually too large, and they are the perfect input here.
A scan is already a photograph of each page — the letters exist as pixels, not as characters. Re-rendering it at a sensible resolution and re-encoding loses nothing you had, because you never had selectable text to begin with. A 40 MB scanned contract becoming a 3 MB one costs you nothing but resolution you were not using.
Worst case: a text document
If your PDF has real text — exported from a word processor, say — compressing it here turns that text into pixels.
The result looks similar and behaves quite differently. It can no longer be selected, searched or copied. Links stop working. Screen readers cannot read it. And the file may well come out larger, because storing crisp text as JPEG images costs more than storing the characters did.
So the tool checks. When it finds a text layer, it says so plainly before you download, and you can decide.
That check is the reason this page exists rather than a simpler one: a compressor that silently destroys searchable text would be the most damaging thing this site could ship, because the damage is invisible until someone needs to search the document.
Choosing a target
- Balanced — for email and upload forms. This is what most files need.
- Smallest — when you are against a hard limit and the document only has to be readable on screen.
- Careful — when fine print, a signature or a stamp still has to be legible.
Each target re-renders from your original file rather than from the previous result, so you can try all three without compounding anything.
What this tool cannot do
Said plainly, because the alternative is more useful than a hopeful answer.
It cannot recompress the images inside a PDF while leaving its text intact. That requires rewriting the document’s object streams, which is a substantially larger piece of work than rasterising.
If you need that — a text-heavy PDF with a few large images, where the text must stay real — Ghostscript is the right tool and runs entirely on your own machine:
gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook \
-dNOPAUSE -dBATCH -sOutputFile=small.pdf large.pdf
It also cannot run OCR, so a scan will not gain searchable text by passing through here.
Other routes to a smaller file
Sometimes the best compression is not compression at all:
- Only part of the document matters → extract those pages
- A cover sheet and some blanks → delete them
- You built it from photographs → compress the images and rebuild it; the result beats compressing the finished PDF
The full picture, including where each route applies: how to make a PDF smaller.
Nothing is transmitted
Rendering runs on PDF.js, the rebuild on pdf-lib, both inside this browser tab.
The PDFs worth compressing are contracts, statements, medical letters and identity documents. The Uploaded readout beside the result measures this page’s own outbound network calls and reads 0 B on success — a measurement rather than a promise.