Compress PDF
Two honest methods, with what each one costs stated.
or drop it here
Nothing is uploaded. Your file never leaves this tab.
What PDF compression is
Two genuinely different operations share the name. Structural compression repacks the file’s internal objects and streams without touching page content. Content-aware compression downsamples the embedded images, which is where a PDF’s bytes actually are — and which needs a library this site cannot legally ship.
What this tool does that others do not
The real number, including when it is negative
On a text document produced by modern software the structural saving is close to zero, because that work was already done at export. On an old or badly generated file it can be worth several percent. Occasionally repacking costs more than it saves, usually on very small documents where object-stream overhead exceeds the gain, and the result shows a negative saving rather than hiding it.
The licensing constraint, stated rather than skirted
Genuinely shrinking a PDF means downsampling its images and subsetting its fonts, and the libraries that do that well — Ghostscript and MuPDF — are both AGPL-3.0. Shipping their code to your browser counts as distribution, which would force this entire site to that licence.
Rasterise offered as the honest alternative
Where the file is mostly scans, the only thing that would shrink it meaningfully is re-encoding those images — which destroys the text layer. That option exists and says what it costs, rather than being presented as "compression".
How to use it
Run the structural pass first
It is lossless and it tells you immediately whether there was anything to take.
Read the figure rather than the label
A saving of 1% on a modern export is the correct answer, not a failure of the tool.
Consider rasterising only if the file is scans
And only if nobody downstream needs to select the text, because they will not be able to afterwards.
Worked examples
Real input, real figures.
A modern text document
You give it
A PDF exported by current word-processing software.
You get
Close to no saving, because the object tables were already compact. The number is shown as it is — a compressor that reported 60% here would be describing something it had not done.
A very small document that grows
You give it
A one-page PDF of a few kilobytes.
You get
The object-stream overhead can exceed the gain, so the result is a negative saving. Keep the original in that case; the figure is there precisely so you can.
Against a service that advertises a percentage
| Aspect | This tool | A typical compressor |
|---|---|---|
| Method named | Structural, or rasterise | Unstated |
| Typical result on a modern PDF | Close to zero, reported as such | A confident percentage |
| When the file grows | Shown as a negative saving | Not surfaced |
| Image downsampling | Not offered, with the licence reason | Offered, on their server |
Guide
Why the saving is usually small, stated honestly
Most of a PDF’s size is its images. Genuinely shrinking one means downsampling those images and subsetting its fonts — and the libraries that do that well, Ghostscript and MuPDF, are both AGPL-3.0. Shipping their code to your browser counts as distribution, which would force this entire site to that licence. We are not doing that for one feature.
What is available permissively is structural: deflating any content that was written uncompressed, repacking the file’s internal objects and, where it is safe, rebuilding the pages into a fresh file so leftovers nothing refers to are dropped. A file with form fields or bookmarks is only repacked, because a rebuild would remove them. On a text document produced by modern software the saving is close to zero, because that work was already done. On an old or badly-generated file it can be worth several percent. The result tells you the real number either way, including when the file gets slightly larger.
Questions people ask
Straight answers. Each one links directly: copy the address after opening it.
Why did my file barely shrink?
Because it was already efficient. Structural compression repacks the document’s object tables and drops orphans; it does not touch image data, which is where the bytes are. If your PDF is mostly scans, the only thing that would shrink it meaningfully is re-encoding those scans at lower quality, which destroys the text layer.
Why not just downsample the images?
Because doing it properly needs Ghostscript or MuPDF, both AGPL-3.0. Using either in a browser tool means distributing their code, which would require releasing all of dsyn under the same licence. Offering a worse result and explaining why seemed better than either relicensing the site or pretending.
My file got bigger. Why?
Occasionally repacking costs more than it saves, usually on very small documents where the object-stream overhead exceeds the gain. The result shows a negative saving rather than hiding it — keep your original in that case.
Nothing leaves your device
Your file is read into this tab and worked on there. It is never uploaded, which is why there is no queue, no daily limit and no account, and why the tool still works with the network switched off.
All 15 PDF tools, and what we do not build