Compress PDF — without uploading it

PDF compression is normally the most upload-dependent tool of the lot, because doing it properly means re-encoding the images inside the file — real work that hosted services run on their own servers. That is why the free tiers cap file size, queue you behind other users, and ask you to hand over documents that are frequently the sensitive kind: scanned passports, contracts, portfolios, tax paperwork.

This one runs Ghostscript compiled to WebAssembly directly in your browser. The engine is roughly a 15 MB one-time download from this site, cached afterwards, and from then on compression is a local operation on your own CPU. Your document is never transmitted — there is no backend here to transmit it to. After the engine has cached once, you can compress with the network disconnected entirely.

Many "compressors" simply rasterise every page into a JPEG and wrap the pictures back into a PDF. The file does get smaller, and it also stops being a document: text is no longer selectable or searchable, copy-paste stops working, screen readers get nothing, and the text turns fuzzy the moment anyone zooms in.

Ghostscript takes the proper route. It re-encodes and downsamples the images inside your PDF, rewrites the internal structure, and removes redundancy — while leaving text as text and vector graphics as vectors. Selectable text stays selectable, searchable text stays searchable, and line art stays sharp at any zoom. Only the embedded raster images lose resolution, which is where nearly all the bulk lives anyway.

• Balanced (150 dpi) — the right default for almost everything. Documents stay comfortably readable on screen and print acceptably.
• Maximum (72 dpi) — for email attachments and upload limits, where getting under a size cap matters more than image fidelity. Screen-legible, not print-quality.
• High quality (300 dpi) — when the PDF will actually be printed. Savings are smaller, but images survive at print resolution.

After it runs you get the before and after sizes side by side, so the decision is based on the real result rather than a promise. If a preset did not help enough, try a stronger one and compare again — each attempt is local and costs nothing but a few seconds.

It depends almost entirely on what is inside. Scanned or image-heavy PDFs frequently drop by 60 to 85 percent on the Maximum preset, because a scan is really a stack of photographs and photographs downsample well. A text-only PDF exported from Word or LaTeX is usually already efficient — there are no big images to shrink, so the saving can be minimal.

When compression genuinely cannot help, the tool says so rather than handing back a file of almost identical size and letting you assume it worked. If a text-heavy PDF is larger than you expect, the cause is often embedded fonts or a scanned cover page rather than the text itself.

Still too large for an email limit? Splitting it into parts with Split PDF is often more practical than compressing further, and converting to grayscale cuts the size of colour scans substantially — both, again, on your device.