PDF Tools
Reduce PDF file size by optimising the document structure and re-encoding embedded photographs, with a quality control so you decide the trade-off. Text is never rasterised.
Overview
This free PDF compressor makes a document smaller without making it worse to read. It re-encodes the photographs and scans inside the file, rebuilds the document structure, and strips the metadata — and it never turns your pages into pictures.
That last point is the difference worth understanding. Some compressors achieve spectacular savings by rendering each page to an image and rebuilding the PDF around it. The file shrinks dramatically and every piece of text in it stops being text: unsearchable, uncopyable, invisible to a screen reader. It is a real saving and a poor trade, and it is usually made without telling you.
So the savings here depend entirely on what is in the file. A scanned document full of high-resolution photographs can shrink a great deal. A text-only PDF exported from a word processor is already close to optimal and may not shrink at all — and when that happens, the tool says so rather than handing back a file that is the same size with a percentage next to it.
Nothing is uploaded. The file is opened, re-encoded and saved by your own browser, which matters given that the documents people most need to shrink are usually the ones they should least like to send to a stranger's server.
Step by step
The file is opened and compressed by your own browser — nothing is uploaded.
Light barely touches the images, strong is for screen and email only. The trade-off is quality of photographs, never of text.
The before-and-after sizes are shown, along with how many images were re-encoded. If the file did not get smaller, it says so.
Background
A PDF is mostly instructions and, sometimes, pictures. The instructions — which glyph to draw where, in which font, what vector shapes to stroke — are compact and already compressed. The pictures are not. In almost any document that is larger than a few hundred kilobytes, the images account for nearly all of it, and the text for a rounding error.
That determines what compression can do. Squeezing a text document harder achieves very little, because there is very little there. Re-encoding a 12-megapixel photograph that has been dropped into a report at full resolution can halve the file on its own. The question is never 'how compressible is a PDF' but 'what is in this one'.
Images inside a PDF are usually stored as JPEG, under a filter called DCTDecode, and the useful consequence is that the JPEG data sits in the file essentially as it would in a .jpg. It can be taken out, decoded, re-encoded at a lower quality or a smaller size, and put back — while every other object in the document, including all the text, is left exactly as it was. That is where an honest compressor gets its savings.
There is also a structural saving, and it is entirely lossless. Re-saving a PDF rebuilds it: objects nothing refers to are dropped, and the accumulated history of a file that has been edited, filled in and re-saved many times goes with them. On a freshly exported document this saves almost nothing. On a form that has been through twenty rounds it can be substantial, because every one of those saves may have appended rather than replaced.
What none of this can do is make an already-tight file smaller. A ten-page text PDF at 90KB is 90KB of genuinely necessary content, and any tool claiming to halve it is either rasterising the pages or lying. Being told plainly that a file cannot be improved is more useful than a download that is the same size.
Reference
Compression is not a property of the tool so much as of the file. These are the things that determine whether there is anything to gain, roughly in order of how much they matter.
Whether the document contains images
The single biggest factor. A scanned document is almost entirely image data and can shrink enormously; a text-only export has nothing substantial to remove and will barely move.
Required
The resolution those images are stored at
Scanners and phone cameras produce far more detail than a page needs. An image at 600 DPI displayed at a fraction of that carries data no viewer will ever show, and reducing it is the cleanest saving available.
Required
The compression level you choose
Light barely touches the pictures and saves modestly. Strong is for documents that will be read on a screen and never printed. The trade is always the quality of photographs, never the quality of text.
Required
The file's editing history
PDFs are often saved by appending rather than rewriting, so a document filled in and re-saved many times carries every previous version inside it. Rebuilding the file discards that, losslessly, and can be a large saving on a well-used form.
Optional
Embedded fonts
Usually small when properly subset and occasionally large when not. Nothing here removes them — a document that cannot render its own text is not smaller in any useful sense.
Optional
What the document is for
A file destined for print needs its resolution; one going into an email thread does not. Compressing to the use rather than to a number is what stops a document being simultaneously too large to send and too soft to print.
Optional
Attachments and embedded files
A PDF can carry other files inside it, and they are counted in its size. If a document is inexplicably large for its page count, this is worth checking in a PDF editor before compressing.
Optional
Whether it is signed
Compressing rewrites the file, which invalidates any digital signature on it. A signed document has to be sent as it is, and the correct fix for its size is usually to split it.
Optional
Who it helps
Mail systems commonly reject anything over 10 to 25MB, and scanned documents pass that easily. Compressing to fit is the immediate need, and splitting the file is the alternative worth considering when quality matters more than convenience.
Scanners default to resolutions intended for archiving rather than for reading. The gap between what a scanner produces and what a recipient needs is where nearly all of the available saving lives.
Portfolios, proposals and photo-heavy reports get sent to clients whose mail servers are stricter than their own. Compressing before sending is faster than discovering the limit after the file bounces.
Applications, tenders and claims frequently impose a size cap with no explanation and no flexibility. Knowing which parts of the file are large is what makes it possible to get under the limit deliberately rather than by trial and error.
The files most in need of compression — statements, contracts, medical scans — are the ones least appropriate to upload to an unknown service. Compressing in the browser removes the question entirely.
Long-term storage rewards smaller files, and rewards keeping the text layer even more. Losing searchability across an archive to save space is a trade that only looks good until somebody has to find something.
Do it properly
Screen reading, email and print need very different resolutions. Choosing the level by what the document is for produces a better result than pushing for the smallest possible number and discovering the pages are soft.
Image re-encoding cannot be reversed — the detail discarded to shrink the file is gone. Keep the source until you have opened the compressed version and looked at the pages that matter.
Open the result and try to select a sentence. If you cannot, the tool rasterised the pages, and you have traded searchability for size without being asked. It is the fastest test there is.
Three 8MB files usually clear a mail limit as easily as one 24MB file and lose nothing at all. Compression is the right answer when the document has to stay whole, not by default.
300 DPI is ample for text and 600 is archival. Scanning at the highest setting available and compressing afterwards throws away detail twice and takes longer than getting it right at the source.
Rewriting the file invalidates the signature, which is the correct behaviour and not something to work around. Send a signed document as it stands, and split it if the size is a problem.
A text-only PDF is already efficient. A tool reporting a large saving on one is doing something to the pages that you would not have agreed to, so no change is often the honest answer.
Avoid these
Using a compressor that rasterises the pages
Check whether the text is still selectable afterwards. Rendering pages to images shrinks a file dramatically and destroys search, copy and screen reader access permanently — a large saving for a loss most people would not have accepted knowingly.
Uploading confidential documents to an online compressor
The files that most need shrinking are scans of statements, contracts and identity documents. Use a tool that works in your browser or desktop software, because anything uploaded is subject to a stranger's retention policy.
Expecting a text PDF to shrink
There is almost nothing in it to remove. If a text-only file is genuinely too large, the cause is usually something unexpected inside it — an embedded attachment, an unsubset font, or a long revision history — rather than insufficient compression.
Compressing hard and then printing
Match the level to the use. Strong compression is fine on a screen and visibly soft on paper, and the mistake only becomes apparent after printing, when the original may already have been replaced.
Deleting the original once it is compressed
Keep it. Lossy re-encoding is one-way, so the compressed file cannot be restored to full quality. As long as the source exists you can compress again at a different level; once it is gone you cannot.
Zipping a PDF to make it smaller
Compress inside the file instead. PDF content streams are already deflated, so wrapping one in a zip typically saves a percent or two while making the attachment more awkward for the recipient to open.
Compressing a digitally signed PDF
Do not. Any rewrite invalidates the signature, and a recipient checking it will see a failure rather than a valid signature. Send the signed file unchanged, and split it if it has to be smaller.
FAQ
The Compress PDF is at the top of this page — free, no signup, nothing uploaded.
Guides
Articles that go deeper than this page has room for.