How PDF fonts and embedding work

Why fonts are a PDF problem at all

A PDF describes exactly where every letter sits on a page, down to precise coordinates, and that precision depends on knowing the exact shape of each character in the font being used. If the computer displaying the file does not have that exact font installed, the reader has to substitute something else, and substitutions can subtly (or not so subtly) shift spacing, line breaks, and the overall look of a page.

Embedding solves this by packaging the actual font data inside the PDF itself, so the file carries everything it needs to render correctly on any computer, regardless of what fonts happen to be installed there. This is why a well-made PDF looks identical everywhere, while a Word document with an unusual font can look noticeably different depending on the reader's system.

Full embedding vs. subsetting

Fully embedding a font includes every character the font defines, which guarantees correctness but adds real file size, sometimes several hundred kilobytes per font. Subsetting embeds only the specific characters actually used in the document — if a report never uses the letter 'Q' in that font, an embedded subset simply will not include it.

Subsetting is the more common default in modern export software, because it gets nearly all of the compatibility benefit of full embedding at a fraction of the file size cost. It is one of the reasons a well-exported, text-heavy PDF stays relatively small even with several distinct fonts on the page.

How this compares to the weight of images

For most everyday office documents, embedded fonts add a modest, largely fixed amount of size — often measured in tens or a couple hundred kilobytes total — while a single unoptimized photo or screenshot can add many megabytes on its own. If a document is unusually large despite having little to no imagery, unusual or heavily embedded custom fonts are worth investigating, but they are a less common culprit than images in practice.

This is part of why Compress PDF focuses its effort on image data rather than font data: images are almost always where the meaningful size reduction opportunity actually lives.

What happens to fonts when a PDF gets compressed

Filemynt's compression targets image streams, not the fonts already embedded in a document — we do not strip, substitute, or re-encode fonts as part of compression. Whatever font handling existed in the file you uploaded — fully embedded, subsetted, or not embedded at all — is preserved through a compression pass rather than altered.

This means compression will not fix a font problem that already existed in the source file, and it will not introduce one either. If text looks subtly wrong after processing a PDF through any tool, the font situation was very likely already present in the original file, not something the tool changed.

Fixing font problems at the source

If a PDF displays with an obviously wrong or generic-looking substitute font on some computers, the underlying fix is to re-export the document from its original source (Word, a design tool, or similar) with a setting like 'embed all fonts' or 'embed subset' explicitly checked, rather than trying to patch fonts inside an already-exported PDF. Most export dialogs offer this option, though it is sometimes buried in advanced or print settings.

Once a properly font-embedded PDF exists, Filemynt's tools — compress, merge, watermark, and the rest — work with that file without touching its font handling, so it is worth getting the export right before any further processing rather than after.

A note on unusual or custom fonts

Brand fonts and custom typefaces used in design files are more likely to run into embedding issues than standard system fonts, simply because fewer computers already have them installed, making embedding more critical for those files rather than optional. If a design proof or branded document needs to look exactly right everywhere it is opened, double-check font embedding specifically before sending, rather than assuming a modern export handled it automatically.