PDFGym
EN

Compress PDF
without losing quality

or drop one here
Compression level
Target size
Choose a PDF file first.

Most PDFs carry weight that has nothing to do with how the pages look: full fonts where a few dozen letters would do, and leftover data from every past save. This page burns that off first, so text documents often shrink a lot with zero visible change, and it always tells you whether a pixel was touched.

The gentle run, step by step

  1. Load the document

    Press Choose a PDF or drop the file onto the page. It stays in this tab on your machine; nothing uploads, which matters exactly when the file is a thesis or a contract.

  2. Aim high, not tight

    Pick the Low level, which targets three quarters of the current size, or type a roomy exact cap in KB or MB. A generous target is what lets the pixel safe methods win the race.

  3. Press Compress

    Font subsetting and the structural rebuild run first, image re-encoding only joins when the gentle pair cannot reach the number, and the full page redraw stays benched unless nothing else got there.

  4. Find out which method won

    Open the Details panel below the result for the full report. A font or structural win means the pages render identically to the original; an image win means pictures changed, and the preview shows by how much. Press Download PDF when you are happy.

How it works describes each method and what it rewrites. If you are chasing a hard number instead of a look, the main compress page takes any exact cap.

The thesis that must stay print sharp

Lena is submitting her doctoral thesis, 140 pages of text, charts and microscopy photos, currently thesis-final.pdf at 48 MB. The university repository accepts 50 MB, so the file fits, but her supervisor wants a copy by email, and the department mailbox stops at 10 MB. The catch: the examination copy must stay sharp enough to print.

She runs it here with a 10 MB cap. Font subsetting strips the unused glyphs from four embedded font families, the rebuild drops years of accumulated save revisions, and the images get a light re-encode to cover the rest. The Details report names the image pass as the winner, so she zooms the preview to full size and checks one chart axis and one photo caption. Both read cleanly. The mailed copy lands at 9.2 MB, and the repository keeps the untouched original.

That split is the professional move: keep the master, send the diet. Never overwrite your only copy with a compressed one when print sharpness matters; storage is cheap, rescanning is not.

Where the bytes live, and which ones are free

The bytes in a finished PDF sit in three places. Fonts ship whole: thousands of glyphs embedded so the pages can draw a few dozen letters. Structure accretes: orphaned objects and old revisions pile up with every save. Images are the only part where smaller can mean visibly worse.

The first two are free savings. Cutting each font to the glyphs the pages actually draw, and copying the live pages into a fresh file without the dead matter, changes nothing on screen at any zoom. There is no quality dial on these operations because there is no quality to trade. That is why a text export can halve without a single pixel moving, while a phone scan of the same pages fights for every kilobyte.

Size breakdown shows the split for your file before you commit: if fonts and structure dominate, expect a big gentle win; if images dominate, expect a negotiation between size and sharpness.

How you know nothing important changed

Two numbers guard every run. The output is reopened and its pages are counted against the original: any dropped page fails the file. Then the non whitespace characters are counted the same way, and anything keeping less than 98% is refused. A refused candidate never reaches the download button; you only ever see files that passed.

What the numbers do not guard is pixels, by design: shrinking images changes pixels on purpose. The pixel judge is you, with the side by side preview. Zoom to full size, find the densest figure or the smallest footnote, and compare. The method name in the report tells you where to look hardest: with a structural or font win there is nothing to find, with an image win the photos deserve a minute of your eyes.

Keep the master file either way. Verification proves the copy is faithful, not that you will never want the originals back.

Traps that quietly cost quality

The classic trap is compressing twice: shrinking an already shrunk file re-encodes the images a second time, and the softness compounds. Always compress from the master. The second trap is a tight cap on a photo heavy file: a 1 MB ceiling on forty phone photos forces the redraw, and no setting makes that sharp. Raise the cap or split the file so each part eats well.

The third trap is judging on a thumbnail. Small previews forgive everything; the flaw shows at full zoom in print. And the fourth is a locked original: a password protected PDF resists the rebuild tricks, so run it through Unlock PDF before you judge what gentle compression can do.

When to skip compression entirely

Some files should not shrink at all. Anything headed for professional print wants the master export, ideally straight from the layout app at press settings; a 150 dpi screen redraw is not a print file. Ask the sender for a fresh export before you squeeze a file that is already small.

And sometimes the kindest cut is deletion, not compression. A portfolio with twelve near identical shots loses nothing when nine go: delete the spare pages first, then compress what remains. Fewer pages at full sharpness always beats more pages gone soft.

Straight answers on quality

Can a PDF really shrink with zero visible change?
Yes, when the weight sits in fonts and dead structure. Those two operations rewrite no pixels, so the pages render identically at any zoom. Whether your file qualifies depends on its insides; the Details report tells you after the run.
How much can the gentle methods save?
It varies widely. Font heavy office exports often drop dramatically; pure scans barely move until images join in. Run once at a roomy target and read the report in Details before deciding anything.
Is the text still selectable afterwards?
Yes. Selectable text stays selectable, and the 98% character check plus the page count check refuse any result that lost pages or text. Even the last resort redraw lays an invisible text layer back over the pages.
Why do photos soften before text does?
Text in a real PDF is drawn from font outlines, which scale cleanly; photos are fixed grids of pixels, and fewer bytes means a coarser grid. That is why photo heavy files need roomier caps than text files.
Should I keep the original after downloading?
Always, when quality matters. Keep the master archived and send the compressed copy. If a recipient ever complains about sharpness, you can reissue from the master in minutes.
Does the compressed file carry a watermark?
No. No watermark, no account, no upload. The download is exactly what the engine produced from your file.

Other tools for careful documents

Inspect, repair, split or convert without leaving the browser. Each one rechecks its output before download, and none asks for an account.

Keep the file sharp