Skip to content
Encoding2026-10-15Begin4 min read

PDF Too Large for Email? How to Choose the Right Compression Quality (3 Local Levels)

A 25 MB attachment cap on email, 10 MB on government portals, 5 MB on job platforms — "PDF too large" blocks more submissions than you'd think. This post explains the two technical routes to compression and how to pick a quality level.

First, diagnose: why is your PDF big?

File size almost always comes down to one of two causes:

  1. Scanned / image-based PDFs: every page is essentially a photograph, and 2–5 MB per page from a 300 dpi scan is normal. Compression pays off most here — often down to under 1/5 the original size.
  2. Text-based PDFs: vector text plus embedded fonts, tens of KB per page. When one of these is big anyway, the culprit is usually high-resolution embedded images or redundant font subsets — compression gains are limited, and structure optimization is the right medicine.

Quick test: try selecting a passage of text in a reader. If it selects, it's text-based.

Two compression routes

RouteHow it worksQualityFits
Lossless structure optimization (useObjectStreams etc.)Rewrites cross-references, drops redundant objects100% losslessText-based PDFs
Lossy rasterizationRenders each page to an image and re-embeds itDepends on quality levelScans / image-based PDFs

Most online compressors take the rasterization route. ToolVault PDF Compress does too — that's where the size savings come from — but it runs entirely in the browser. Nothing gets uploaded, which for documents containing ID cards or contracts is a categorical difference.

Choosing among three quality levels

With an A4 document as the reference (content varies a lot; treat these as orders of magnitude):

  • High quality (≈150 dpi, quality 0.9+): printing, archiving, formal submissions. Visually indistinguishable from the original; scans typically compress to 1/2–1/3.
  • Medium quality (≈120 dpi, quality 0.8): screen reading, email attachments. The smallest type softens slightly — acceptable in most cases, usually down to 1/4–1/5.
  • Low quality (≈96 dpi, quality 0.6): when "just openable" is enough. Smallest size; fine print starts to blur. Never use this for anything meant to be printed.

One discipline worth adopting: try medium first, open the result and check the critical pages (smallest text, charts, seals), and step up a level if it fails. Compression is irreversible — always keep the original.

Three truths about compression

  1. Text-based PDFs barely compress: vector text is already small, and rasterization can even enlarge the file (it turns text into images). What those files need is stripping the redundant embedded fonts.
  2. "Compressed bigger" is normal: rasterizing an already-optimized PDF can cost more in image re-encoding than it saves. A tool should show the before/after sizes honestly instead of pretending success.
  3. Black-and-white scans compress better in grayscale: color rasterization is waste on monochrome content. Some tools offer a color-mode option; choosing grayscale for things like scanned ID photocopies halves the size again.

Command-line alternative

With a server or local environment available, Ghostscript is the standard answer for restructuring compression:

gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -o out.pdf in.pdf

The three presets map to the levels above: /screen (96 dpi), /ebook (150 dpi), /printer (300 dpi). In scripted pipelines it beats the browser tool.

Implementation notes

Key trade-offs when building in-browser compression: rasterization render scale and JPEG quality are two independent knobs, but users shouldn't be forced to understand both — collapsing them into three presets (internally mapped to scale + quality combinations) is the better product decision. Another trap is hard canvas size limits: rendering a large-format page at high scale silently yields a blank image in some browsers, so you must guard the longest edge and step the scale down. Show the before/after sizes and page counts in the results area, and display inverted results like "-3% compression" honestly — pretending every run succeeds damages trust more than an honest failure does.

Bottom line

Over-limit scans: medium rasterization (ToolVault PDF Compress, local, no upload); for print use, go high. Text-based PDFs won't compress much — don't waste expectations on them. With a CLI available, use Ghostscript. And keep the original, always.