Compress a PDF without wrecking it

A PDF is big for one of a small number of reasons, and they call for opposite treatments. Guessing wrong is how people end up with a fuzzy file that is also no smaller.

• Photographs. A brochure with twenty full-page images is large because of the images, and compressing them is the entire job.
• A scan. Every page is one big picture, even the ones that look like text. These compress dramatically and are the easiest win.
• Embedded fonts. A document using six typefaces can carry several megabytes of font data. Subsetting helps; re-compressing the page does nothing.
• Vector artwork. Maps, CAD exports and complex charts can be tens of megabytes of pure geometry, with not a single image to squeeze.
• Nothing in particular. A 400-page text document is legitimately large. There is no trick here.

A quick way to tell: open the file and try to select a line of text. If you can, the page is real text and rasterising it will hurt. If you cannot, it is a scan, and it will compress beautifully.

This surprises people, and the cause is simple. Many compressors work by rendering every page to an image and rebuilding the PDF from those images. On a scan that is a huge win. On a page of crisp text it is a disaster twice over: the text becomes a photograph of text — fuzzy, unselectable, unsearchable — and a photograph of a page of text is often larger than the instructions that drew it.

A 40 KB text page can easily become a 400 KB image. Run that over a whole document and you have a file ten times the size with none of the searchability.

1. 1Try plain compression firstRecompressing the embedded images and streams without rasterising anything is lossless for the text. Often that alone is enough, and nothing degrades.
2. 2If the file is a scan, rasterise deliberatelyIt is already images, so there is nothing to lose. This is where the dramatic reductions live — 80% or more is normal.
3. 3Check the size before you accept the resultCompare against the original. If it did not get meaningfully smaller, the size is coming from something that compression cannot touch, and more aggressive settings will only cost you quality.
4. 4Look at the text at 100%, not at the thumbnailRasterisation damage is invisible in a page thumbnail and obvious at full size. Zoom to a paragraph and look at the letter edges.

• Too large to email? Splitting the document is often better than degrading all of it. Sending chapters 1–3 at full quality beats sending everything blurred.
• Only some pages needed? Extract those pages. A 2 MB extract of the relevant section is a better file than a 4 MB compression of the whole thing.
• Made it yourself? Re-export from the source with images downsampled at export time. That always beats squeezing them afterwards, because the original never has to be thrown away twice.
• Scanned it yourself? Rescan at 200–300 DPI in greyscale. Most people scan documents at 600 DPI in colour and then wonder why the file is enormous.