PDF300

Compress a PDF Without Losing Quality: What DPI, JPEG Quality and Downsampling Actually Do

Updated · 4 min read · By

The situation

Every compressor promises "no quality loss". Then the 120 MB portfolio becomes 8 MB and the renders look like they were faxed. Or the opposite: the "lossless" option runs and the file drops from 120 MB to 117 MB.

Neither tool lied, exactly. They used the word quality to mean different things. This guide explains what is actually being changed so you can choose settings on purpose.

Two kinds of compression inside a PDF

Lossless rewrites the file without changing a single pixel or glyph:

Typical saving: 5–30%. On a scan or a photo-heavy export: close to nothing, because the images were already JPEG.

Lossy changes the images:

Typical saving: 50–95%. This is where the big numbers come from, and this is what "without losing quality" refers to: lossy compression tuned so that you will not notice the loss at normal viewing. Whether you notice depends on what you look at and how closely.

How much resolution you actually need

Resolution in a PDF is about pixels per inch at the size the image is placed on the page. A 3000-pixel-wide photo placed 3 inches wide is 1000 ppi. Downsampling it to 200 ppi throws away 96% of its pixels and changes nothing visible on the page, on screen or in print.

Use Image resolution that is enough
Reading on screen (laptop, tablet) 96–150 ppi
General office printing 150–200 ppi
Small print, footnotes, handwriting, stamps 200 ppi minimum
OCR input, archival of text pages 300 ppi
Fine line art, maps, engineering detail 300–600 ppi, or keep it vector

Printers put dots at 600–1200 dpi, but a photograph needs far fewer pixels than the printer has dots; 200–300 ppi is the long-standing print standard for images.

JPEG quality

JPEG quality is a dial, usually 0–100. Roughly:

JPEG is designed for photographs. Line art, diagrams and text screenshots develop halos even at high quality settings; keep those as lossless (Flate/PNG-style) images or, better, as vector content.

Color to grayscale

A color image has three channels; grayscale has one. Converting typically cuts an image's size by 40–60% and is invisible on documents that were black-and-white to begin with — which includes most scanned contracts, forms and letters. It is the single largest "free" saving on a scan that does not have colored stamps or highlights that matter.

Vector content is never downsampled

Text, lines, shapes and fonts are instructions, not pixels. Compression settings do not touch them; a drawing exported from CAD stays sharp at any zoom after compression. If a tool blurs your text, it rasterized the page — that is a different (and bad) operation, not compression.

How to judge the result

  1. Open the original and the compressed file side by side.
  2. Zoom both to 100%, then 200%, on the page with the smallest text and on one photograph.
  3. Read a number. Numbers are where errors cost money.
  4. If it will be printed, print one page.
  5. If it had an OCR layer, search for a word.

If it passes all five, the quality loss is real but irrelevant. That is the correct target.

Choosing settings on purpose

A target-size tool such as PDF300 runs this trade-off backwards: given the size you must hit, it finds the mildest settings that get there, which is usually a better outcome than guessing a preset and hoping. Either way, the five-step check above is what tells you whether the result is acceptable.

FAQ

Is compressing a PDF twice worse than once?

Each JPEG re-encode loses a little more; the loss compounds. Two passes at quality 80 are worse than one pass at quality 70. Always compress from the original, never from a previous compression.

Why does the "lossless" option barely shrink my scan?

Because a scan is already a JPEG. Lossless methods can only tidy the container around it. Real savings on a scan come from downsampling and grayscale.

Will my drawings or charts get blurry?

Not if they are vector — compression does not touch them. Raster backgrounds (aerials, scanned base maps) do get downsampled; keep those at 150–200 ppi for review sets and confirm requirements for sealed deliverables.

Written by LeDeuxions, an architect who built PDF300 after a 300 MB portal limit kept rejecting drawing sets. Figures quoted from third parties link to their source; sizes and reductions described as "typical" are estimates, not guarantees.