Compress PDF Without Losing Quality — What Works

GN PDF · February 2, 2026 · 10 min read · Tutorial· Compress

Where does PDF "size" actually come from?

A 25 MB PDF is rarely 25 MB of text. The size is almost always three things, in this order of magnitude:

Embedded images at full resolution — phone-camera photos, scanned pages, screenshots — even if they appear small on the page.
Embedded fonts at full glyph coverage. A single font face used for one character can still cost 200-400 KB if the full set is embedded.
Uncompressed object streams — the PDF's internal data that wasn't subjected to deflate/flate compression on save.

If you don't address image bloat, you cannot meaningfully compress a PDF. Everything else is at most a 10% saving.

Three legitimate compression strategies

1. Downsample and recompress images

The biggest win. Take every embedded image, downscale it to the resolution it actually needs (screen reading = 100-120 DPI; print = 300 DPI), and re-encode it as JPEG at 75-85% quality. This is what our Compress PDF tool does, and it's where you get the dramatic 60-80% shrink on scan-heavy files.

2. Subset fonts

Instead of embedding all 6,000 glyphs of Noto Sans, embed only the 50-200 glyphs actually used in the document. A good PDF writer (including pdf-lib) does this by default; older PDFs created by Word or InDesign often do not.

3. Re-stream with deflate

Re-saving a PDF through any modern PDF library forces all object streams through deflate compression. This typically nets a 5-15% reduction on already-optimised files, more on PDFs created by old tools.

The strategies you should AVOID

Aggressively flattening to a single image per page. Yes, this shrinks file size dramatically, but it kills selectable text, hyperlinks, and bookmarks. Use only if file-size is the only thing that matters (e.g., a 5 MB UPSC upload limit).
Lowering DPI below 100 on print-bound PDFs. The result will look fine on screen and unreadable on paper.
Removing metadata to save bytes. The savings are negligible and you lose useful information like document title and author.

A real example

I took a 22.4 MB lease agreement scanned at 600 DPI. The text was perfectly readable at 200 DPI. Running it through Compress PDF on the "Balanced" preset:

Output size: 2.7 MB (88% reduction).
Text remained crisp when zoomed to 200%.
Embedded images (a photo of the property) were re-encoded at JPEG 78% but were visually indistinguishable on screen.

Now the same file through "Maximum compression":

Output size: 980 KB (96% reduction).
Text was readable but slightly softer at 200% zoom.
The property photo lost some fine detail — acceptable for an attachment but not for archival.

What about "compress without losing quality"?

Pure lossless PDF compression — without re-encoding any image — typically gives you a 5-20% reduction. That's it. Anyone claiming a 90% lossless reduction is either using a misleading definition of "lossless" or compressing a deeply unoptimised source.

Practical recipe by use case

Goal	Best setting	Typical reduction
Email attachment	Balanced	50-80%
UIDAI / GST / 2 MB portal	Maximum	80-95%
Print-quality archival	High	10-30%
Sharing with a designer	High	10-30%

Closing

Compression isn't magic, but with a few sensible choices you can hit any reasonable size target without compromising readability. Try Compress PDF with your worst-offender file and see what you get.