"Compress PDF" tools promise dramatic reductions, but the physics is simple: a PDF is only as heavy as its images, fonts, and structure, and only one of those can shrink for free. Here's what lossless compression actually buys (measured, not marketed), why scans are the files that shrink 80 percent while text PDFs barely move, and how to pick settings that hit your size target without wrecking the document.
Three places. Images: every photo, scan, and screenshot, stored at its original resolution. Fonts: embedded font files, which can run from a few kilobytes (a subset) to several megabytes (full families). Structure: the objects, cross-reference tables, and revision history that hold the file together — meaningful only in files saved by inefficient writers.
The proportions decide your strategy. A text-only PDF is mostly structure and subset fonts — small, and nearly incompressible. A scanned PDF is image stacked on image — enormous, and compressible, because the images can be recompressed. That single distinction explains almost every "why did my file barely shrink?" complaint.
PDF 1.5 introduced object streams: small objects get packed into compressed containers instead of sitting individually in the file. Rewriting a PDF with them (plus dropping unused objects) is lossless — every glyph and pixel stays byte-identical.
The honest number: on a 10-page, 400-line text document built for testing, the rewrite trimmed 2.5 percent — 21,865 bytes to 21,321. That's the realistic expectation for documents that were already saved efficiently. Files from bloated writers (some scanners, some office suites) can carry more dead weight and shrink further; files saved by modern tools shrink less. When a service advertises "lossless, 80% smaller," images are being recompressed — which is lossy, whatever the label says.
The other approach redraws every page — render it, then encode it as a JPEG — and builds a fresh PDF from those images. This is where scanned documents collapse: a 300 DPI A4 scan carries about 2480 × 3508 pixels per page, and rerendering at 150 DPI replaces it with a 1240 × 1754 JPEG — roughly a quarter of the pixels, at stronger compression. For a file that will be read on a screen, 150 DPI at JPEG quality 70–80 is the workhorse setting.
The trade is real, though. Once a page is an image, its text stops being selectable and searchable, form fields stop being fillable, and fine detail is capped at whatever DPI you chose. It's a one-way door for the content — keep the original.
| Method | Keeps quality? | Keeps selectable text? | Typical result |
|---|---|---|---|
| Object streams (lossless) | Byte-identical | Yes | Small gains — 2.5% measured on a text test doc; more on bloated files |
| Rerender scans at 150 DPI | Visually fine on screen | No — pages become images | The dramatic cuts, on scan-heavy files |
| Rerender text PDFs | Looks similar | No | Often grows — vectors were already compact |
| Font subsetting | Identical rendering | Yes | Large only when full fonts are embedded |
Email is the usual deadline. Gmail caps attachments at 25 MB; many corporate gateways are stricter, commonly 10–20 MB. Portals and application systems often set their own (job boards around 5–10 MB is typical). When a rerendered scan still won't fit, trimming to the pages the recipient actually needs — page ranges like 1-5,12 — beats another round of quality loss.
Lossless repack or 150 DPI rerender, with before/after sizes reported every run — free, private, in your browser.
Open Compress PDF →Related moves: PDF to JPG pulls individual pages out as images, JPG to PDF rebuilds scans into a document first, and PDF Add Text stamps a date on pages before you send them. The compression math stays the same wherever you do it.
Lossless is safe and modest — a few percent, byte-identical content. The big numbers come from images, so rerender when your file is scans and the destination is a screen: 150 DPI, quality 75, keep the original. Pick by content, check the reported sizes, and stop as soon as it fits.
Use lossless compression: the file is rewritten with PDF 1.5 object streams, which pack small objects into compressed containers. Every pixel, glyph, and vector stays identical. The honest ceiling: measured on a 10-page, 400-line text test document, object streams trimmed 2.5 percent (21,865 to 21,321 bytes). If someone promises 80 percent smaller and lossless, images are being recompressed — that's lossy, whatever the marketing says.
A scanned page is a full-resolution image — a 300 DPI A4 scan is about 2480 by 3508 pixels per page — while a text page is a few kilobytes of vector instructions. Rerendering a scan at 150 DPI replaces each page with a 1240 by 1754 JPEG at roughly a quarter of the pixels, which is where the dramatic cuts come from. Rerender a text PDF and it usually grows, because you replaced compact instructions with an image.
For on-screen reading, 150 DPI at JPEG quality 70 to 80 is the workhorse setting — an A4 page becomes 1240 by 1754 pixels, sharper than most displays show. Go 200 to 300 DPI if the document will be printed or needs deep zooming. Below about 100 DPI, body text starts to soften noticeably.
Attachment limits vary by provider — Gmail caps at 25 MB, and many corporate systems are stricter at 10 to 20 MB. When a file won't fit, the usual fix is rerendering a scan at 150 DPI; if it's still over, drop the page range to just the pages the recipient actually needs, or share a link instead of an attachment.
Lossless compression leaves content byte-identical — no damage, no visible change. Rerender compression flattens each page into an image: visually similar at a sane DPI, but text stops being selectable and searchable, forms stop being fillable, and fine detail is capped at the DPI you chose. Keep the original when you rerender.