Compress PDF

Two modes, two different trades

Most “compress PDF” tools present a single slider and do not explain what it costs. There are really two distinct operations, and which one you want depends entirely on what is inside your file.

Keep text sharp (lossless)

The document is parsed and rewritten with compressed object streams, and metadata - title, author, producer, keywords - is stripped. Page content is untouched.

• Text stays selectable and searchable
• Image quality is unchanged
• Typical saving: 5–20%, sometimes zero

This is the safe option, and the right one for a text document, a report or an invoice exported from software.

Shrink hard (image-based)

Every page is rendered to a bitmap at a chosen resolution, compressed as JPEG, and placed into a new PDF at the original page dimensions.

• Text becomes part of an image - not selectable, not searchable, not accessible to screen readers
• Typical saving on scans: 70–90%
• Page size and layout are preserved exactly

This is the right option for a scanned document, which is already just images, and the wrong option for anything where the text matters as text.

Why file sizes vary so much

What dominates the file	Lossless saving	Image-mode saving
Scanned pages	Almost none	70–90%
High-resolution photographs	Almost none	60–85%
Text with embedded fonts	10–20%	40–70%, and you lose the text
Vector charts and diagrams	5–15%	Often worse - vectors compress better than their rasterisation

That last row is worth noting. Rasterising a vector chart makes it larger, not smaller, as well as blurrier. If your PDF is mostly diagrams, image mode is the wrong tool.

Accessibility is a real cost

Rasterising a document removes it from anyone using a screen reader. It also breaks text search, copy-paste, translation tools and any downstream automated processing.

For personal archiving that may be an acceptable trade. For anything you send to another person, a client, or a public body, it usually is not - and in several jurisdictions, documents published by public sector organisations are legally required to be accessible.

If you need both small and searchable, the answer is OCR rather than compression: run the scan through optical character recognition to add a hidden text layer, then compress the images beneath it. That is a Phase 2 capability here.

Metadata removal

Lossless mode clears the title, author, subject, keywords, producer and creator fields. These routinely leak more than people expect - the original author’s name, the software and version used, and sometimes an internal file path from the machine that created it.

Note this removes document metadata. It does not remove content from within the pages themselves. Text hidden behind a rectangle is still text, and it is still there.

If it is still too large

Try these in order: strip pages you do not need with the PDF splitter; use image mode at 120 dpi and quality 70; drop to 96 dpi. If a hard limit still defeats you, splitting the document into parts is usually more honest than degrading it to illegibility.