Compress PDF
Shrink a large PDF, honestly.
- Deleted from our server immediately
- Free, no sign-up
This one needs our server. Unlike the rest of our tools, this file is sent to us to be processed, then deleted the moment your download is ready — never written to disk, never backed up, never looked at.
Related tools
All pdf toolsQuestions
Why does this tool need your server when the others do not?
Because the size of a PDF is almost entirely the images inside it, and shrinking those means decoding and re-encoding real image data with a real codec — which no browser exposes to a web page. Every other tool on this site runs on your device precisely because it can. This one cannot, so your file is sent to us, processed in memory, and discarded the moment your download is ready. It is never written to disk, never backed up, and never read by a person.
How much smaller will my PDF get?
It depends entirely on what is in it. A scanned document — pages that are really photographs — commonly drops by 70 to 90 percent, because a scanner writes far more resolution than anyone reads on screen. A PDF made of text and vector graphics may not shrink at all, because there are no images to re-encode and the text was already compressed. The tool tells you the actual figure rather than a promise.
What if it cannot make my file smaller?
You get your original file back, and the page says so plainly. A compressor that returns a file the same size while implying it did something is worse than no compressor, so this one refuses to pretend. That outcome usually means the PDF is text and vectors rather than scans, or that its images were already compressed harder than we would compress them.
Will the text still be selectable and searchable?
Yes. Only the images inside the document are touched — text, fonts, links, form fields and bookmarks are carried through unchanged. If the PDF is a scan with an OCR text layer, that layer survives too. What changes is the resolution and encoding of the pictures, which is where the bytes actually are.
Will diagrams and line art be ruined?
No, because they are deliberately skipped. JPEG compression rings around hard edges, so a chart or a scanned line drawing can encode smaller and still look worse. The tool counts the distinct colours in each image and leaves anything that looks drawn rather than photographed exactly as it was, even on the strongest setting.
Is there a file size limit?
20 MB, which covers the overwhelming majority of documents people need to shrink — and a PDF above it is usually a scan at a resolution no one asked for. The limit exists because the whole file is held in memory rather than written to a temporary file on our server, which is what lets us say it never touches a disk.
How it works, and more about compress pdf
How it works
- Choose your PDF. Drag it onto the drop zone or click to browse. Files up to 20 MB.
- Pick how hard to compress. Balanced suits almost everything. Choose Smallest file when an upload limit is the problem, or Best quality when the document is going to print.
- Download the result. You are told exactly how much was saved before you download. If we could not make it meaningfully smaller, we say so and give you back the file you started with.
Where the size actually is
A PDF that is too big is almost never too big because of its text. Text and vector graphics are compact and already compressed — a hundred pages of them is a few hundred kilobytes. The weight is the images, and in a scanned document the images are the document: every page is a full-bleed photograph, written at whatever resolution the scanner defaulted to, which is routinely 300 or 600 DPI. Nobody reads a contract at 600 DPI. Most people read it on a laptop at about 110.
So compressing a PDF properly means decoding those images and re-encoding them smaller. That is the whole job, and it is why this is the one tool here that needs a server: a browser will happily decode an image for display, but it does not expose the encoder a web page would need to write the result back into the file.
Why this one is not on your device
Every other tool on this site runs entirely in your browser, and we would rather say that loudly than quietly make an exception. This page is the exception. Your PDF is sent to our server, held in memory while it is processed, and discarded the moment your download is ready. It is never written to disk, never included in a backup, never read by a person, and never used to train anything — there is no database in the service and no directory it could be written to.
If a document is sensitive enough that you would rather it did not leave your machine at all, that is a reasonable position and we would rather you acted on it. Everything under PDF tools — merging, splitting, rotating, deleting pages — runs on your device and always will.
What the three settings actually change
Each one sets a target resolution and a JPEG quality. Smallest file reduces images to 96 DPI, which is screen resolution: sharp on a monitor, visibly soft in print. Balanced targets 150 DPI, the usual scanning default, which prints acceptably and is the right answer for almost every document. Best quality targets 220 DPI, close enough to the 300 most scanners produce that telling them apart takes a loupe.
Note that these are ceilings, not targets. An image already below the threshold is left at its own resolution rather than being upscaled — compression here can only ever remove detail, never invent it.
What is deliberately left alone
Three kinds of image are skipped no matter which setting you choose, because compressing them would cost more than it saves.
Line art, charts and screenshots. JPEG works by discarding high-frequency detail, which is exactly what a hard edge is made of, so it rings around text and diagram strokes and leaves a grey halo. A chart can encode smaller as JPEG and still be a worse file than it started as. Anything with few enough distinct colours to have been drawn rather than photographed is left untouched.
Specialist codecs. JPEG 2000, JBIG2 and CCITT Group 4 are already better at their particular jobs — JBIG2 in particular compresses scanned black and white text far harder than JPEG ever will. Replacing them would be a downgrade.
Anything that came out bigger. Every re-encoded image is compared against the original and kept only if it is genuinely smaller. The same check runs on the finished file. This is what makes the tool safe to point at anything: it cannot hand you back something larger than you gave it.
When it will not help
If your PDF is text and vector graphics — anything exported from a word processor, a spreadsheet or a design tool rather than scanned — there are no images to re-encode and the saving will be a percent or two at best, from repacking the file’s internal structure. The tool will tell you that plainly and hand back your original rather than dress up a rounding error as a result.
The same is true of a PDF someone has already compressed. Quality lost to JPEG does not come back, and running it through a second time removes a little more detail for a saving that is usually not worth it. If you have the original scan, compress that instead of the compressed copy.
What survives
Only the images are touched. Text stays selectable and searchable, fonts are carried through unchanged, and links, bookmarks, form fields and page structure are preserved. If the file is a scan with an OCR text layer behind the images, that layer is not an image and is left alone — so a searchable scanned document is still searchable afterwards.
Transparency survives too. Soft masks are stored as separate objects in a PDF rather than inside the picture data, so they are left in place rather than flattened, which is the usual way a compressed PDF ends up with black boxes where something used to be see-through.