How to Compress a PDF Without Any Software

Modern browsers run the PDF engine in JavaScript and WebAssembly. Your file is read locally, embedded images are re-encoded at the selected quality, content streams are rebuilt, and the smaller pdf file is downloaded — no installer, no upload, no watermark.

One honest correction: the browser is itself software. What you avoid is installing, registering or licensing anything on a machine that may not belong to you.

Three contributors dominate the size of most files. Embedded images — photos, scans, screenshots — usually account for sixty to eighty percent of the weight. Fonts come second: every embedded typeface adds hundreds of kilobytes unless it is subset to the characters actually used.

The third contributor is invisible: metadata, redundant streams and edit-history residue that accumulate every time a file is saved, re-saved and forwarded.

The preset controls image re-encoding. High keeps images near-original, balanced down-samples them moderately, and smallest aggressively recompresses them. Text and vector lines are rebuilt with negligible loss at every level — which is why a 40-page text contract compresses far more than a six-page brochure of photos.

The right target size is set by the destination, not by a round number. Mail servers, chat apps and portals each have their own ceiling, and each rejects a file that sits just above it.

Aim ten to twenty percent under the limit. A file sitting at 24.8 MB against Gmail's 25 MB still fails mid-transfer, and nobody explains why.

The composition of the file decides what compression can promise. Text and vector pages shrink dramatically because their weight is mostly instructions, not pixels; image-dominant pages hit an image floor quickly.

A scanned page is already a compressed image. Recompressing it produces a smaller but uglier page, not a small page — the file shrinks a little and the text gets visibly softer.

Two honest fixes exist. Re-scan at 150–200 DPI instead of 300, or OCR the text and rebuild the document so the image weight is replaced by text weight. A rebuilt scan can drop by half or more.

The route should match the machine you are stuck with. On a shared office computer or a borrowed laptop, the browser is the only option that starts without asking anyone's permission.

The full run takes under a minute for a 30 MB file. Open the compressor in the browser, drop the file in, choose the preset by destination, run it, and compare the output against the original in your downloads folder.

Kavita Rao, a marketing manager in Bengaluru, faced exactly this: a 31 MB product brochure PDF and a client server that capped at 10 MB. One pass at balanced quality produced a 7.2 MB file; the client opened it on the first try and the print shop later used the high-quality original kept aside.

Three checks take thirty seconds: confirm the page count matches the original, zoom to 100% on an image-heavy page and confirm it still reads cleanly, and open the file once in a fresh viewer to confirm no fonts went missing.

If anything looks soft, re-run at the next quality level up. A re-send costs a minute; an unreadable attachment costs the day.

Below a certain size, quality losses start showing where they hurt — in printed copies, in tiny text, in charts the client tries to zoom. Compression has a point of diminishing returns, and a good rule is to stop while the file still prints well.

Keep the high-quality version as the master and compress only the travelling copy. A local-processing compressor like Crafex's keeps even that travelling copy on your device — for confidential quotes and cost sheets, that is the setting that matters as much as the size.

The destination decides your target, not the source file. Gmail and Outlook both cap attachments near 25 MB; the sternest portals — courts, GST filings, tender submission systems — frequently halve that. Aim for one of three bands: under 25 MB for any email, under 10 MB for the picky mail servers, and under 5 MB when a portal announces a hard cap.

When a client's account is the real ceiling, ask or guess conservatively and test with a self-send. A file that arrives on your own inbox arrives on theirs; a file that bounces back from yours was never going to land at all.

For scanned paper — contracts, past invoices, KYC sets — the scan resolution you choose upstream matters more than any compressor. Re-scanning a bad 32 MB scan at 200 DPI greyscale often halves the trouble before compression begins.