You export a three-page document, and the PDF comes out at 28 MB. No scans, no videos, no images except maybe a logo — and yet it will not fit in an email. Before you reach for a compressor, it is worth knowing where the bytes actually live, because the fix depends on the cause.
The fonts nobody thinks about
Every PDF embeds its fonts so it looks the same on every machine. A full font family with bold, italic and bold-italic is four files; at 300 KB each that is over a megabyte before you have written a word. Documents using many typefaces (reports with a serif body, a sans headline and a monospace code block) can carry several megabytes of embedded fonts.
A structural compressor subset-encodes the fonts — it keeps only the characters you used — which is why Light compression on a Word export typically saves 20–40% with zero visible change.
Images are the real weight, usually
Any image in the PDF is stored inside it. A photo pasted into Word at full quality and cropped small is still stored at its original resolution and size — the crop is just a window. A 4000×3000 photo from a phone is 3–8 MB all by itself, even if it prints as a thumbnail on page two.
Two common duplication traps:
- The same image pasted repeatedly — the company logo on every page of an invoice template, stored once per page instead of once per file.
- The same image at full quality plus a thumbnail copy — many editors store a preview plus the original.
Export settings decide most of it
| Where it came from | Typical cause | What to do |
|---|---|---|
| Word / Google Docs export | Unsubsetted fonts, full-size images | Light compress, then delete unused pages |
| Print-to-PDF | Rasterised pages, no structure | Strong compress, expect quality loss |
| Scanner | Uncompressed bitmap pages | Balanced or Strong; re-render, not restructure |
| Photoshop “Save for web” | Already optimal | Leave it alone — further compression gains ~0 |
How to check what is inside yours
Open the file in a PDF viewer and look at Document Properties (Ctrl+D in most). The page count, PDF version and producer (Word, Distiller, Acrobat, a scanner) tell you the likely cause:
- Producer: Word/LibreOffice → fonts and images; Light compression will do most of the work.
- Producer: scanner/Distiller → images; you need Balanced/Strong re-rendering.
- Pages with no images but huge size → fonts and structure; subset fonts with Light.
The fix, in order
- Delete first, compress second. Blank pages and duplicate inserts are pure dead weight; removing three of twelve pages is a 25% saving for free. Use Delete PDF pages.
- Replace full-quality images at the source with ones sized for the document if you can re-export the original. This beats any downstream compression.
- Run Compress PDF Light for text documents. It deduplicates fonts, re-packs structure and re-compresses embedded images losslessly where possible. No visible change, usually 10–40% smaller.
- Use Balanced for mixed documents (text plus photos). It re-renders images at a printable resolution around 150 DPI.
- Strong only for scans you need to email and do not need in print — it re-renders the whole page and throws away any text layer.
- Still too big? Split it. Split PDF and send only the pages the recipient needs. The smallest possible attachment is the one you never send.
Things that do not help
- Saving and re-saving in Word. Each cycle can re-embed fonts again; it usually grows the file, never shrinks it meaningfully.
- Zipping. Already-compressed data does not zip. A 28 MB PDF in a ZIP will still be about 28 MB.
- Renaming the file. The bytes do not care what the file is called.
- Strong compression on a text PDF. Softens the glyphs, can rasterise text, and saves little because there is no image layer to squeeze. Light is the right tool for text.
Why do it on your own machine
A 20 MB PDF is often the document you cannot email around: a contract, a statement, a medical form. Uploading it to a compression website means a copy of that document sits on someone else’s server. Everything linked here runs the compression in your browser — the file is read, rewritten and written back on your own device, and nothing is uploaded. That matters more the more sensitive the document is.