Compress a PDF without uploading it

Scanned documents are heavy because every page is a photograph. Compression re-encodes those images at a lower resolution and quality, which is usually the difference between a 24 MB email attachment and a 2 MB one. Everything happens in this tab: choose the file, choose how aggressive you want to be, get the smaller file back.

• Two presets: screen for the smallest file, ebook for a readable middle ground
• Measured on a public benchmark corpus, not estimated
• Never returns a file larger than the one you gave it
• No upload, so no size cap imposed by a server

Frequently asked questions

How much smaller will my PDF get?

Measured on our benchmark corpus: a 10-page 300 dpi scan went from 3.95 MB to 0.17 MB at the screen preset, a 96 percent reduction, and to 0.47 MB at ebook. A 150 dpi scan, which has less redundancy to throw away, dropped 87 percent and 61 percent. The numbers and the corpus that produced them are in the repository, so you can rerun them.

What about a PDF exported from Word?

It will come back unchanged. Compression here works by re-rendering each page as an image, which is a large win on a scan and a loss on a document that is already vector text - on a 20-page Word export the re-rendered version was over a hundred times larger. Rather than hand you that, the tool detects it and returns your original file untouched.

Will compression ruin the quality?

The screen preset is visibly softer at full zoom and is meant for reading on a display or attaching to an email. Ebook keeps noticeably more detail at roughly three times the file size. Start with ebook and drop to screen if it is still too big for what you need.

Can I compress a PDF that contains confidential data?

That is the reason this tool exists. Payslips, medical records, contracts and tax returns never leave the browser, so there is nothing sitting on someone else's disk afterwards.

Guides and benchmarks

• How small a PDF actually gets, from 50 KB to 1 MB - Five test files, both presets, and which size targets each one clears. Measured in the browser, with nothing uploaded at any point.
• What PDF compression actually does, measured - Every PDF site claims "up to 90% smaller" and none of them show their corpus. Here are the numbers for this one, the documents they came from, and how to reproduce them.
• OCR that looked like it worked - Five bugs in a browser OCR pipeline, every one of which produced a plausible success instead of an error. Measured before and after, with the code paths named.