unsubsetter 0.1.0

Re-embed full (non-subset) fonts in PDFs.

unsubsetter

Re-embed full (non-subset) versions of fonts in PDFs.

PDF generators often subset embedded fonts, keeping only the glyphs a document uses and marking the font's PostScript name with a six-letter ABCDEF+ prefix. Some preflight checkers treat subsetted fonts as "not embedded" even though they are. unsubsetter swaps each subset for the complete font found on disk. It was built to get a book past Amazon KDP's preflight check, but applies to any PDF that needs non-subsetted embedded fonts.

It handles CID TrueType and CID CFF fonts. Other font types are detected and reported, but left unchanged — see Limitations.

Install

Install the unsubsetter command into an isolated environment on your PATH:

uv tool install unsubsetter

Or with pipx:

pipx install unsubsetter

From source

git clone https://github.com/saggingmeniscus/unsubsetter
cd unsubsetter
uv tool install .

Usage

Inspect (default — no writes):

unsubsetter book.pdf

Fix (writes book.unsubset.pdf by default):

unsubsetter --fix book.pdf

Filter to specific fonts:

unsubsetter --fix --only Garamond,Helvetica book.pdf

With visual verification (renders N random pages and pixel-diffs them):

unsubsetter --fix --verify-visual 10 book.pdf

Example: preparing a PDF for Amazon KDP

KDP's preflight check is the use case this tool was built for — it can reject a PDF whose fonts look un-embedded, which subsetting may trigger. A careful pre-upload pass:

1. Inspect (default mode, no writes):

   unsubsetter interior.pdf
   

   Confirm the plan covers the font KDP flagged. Resolve any surprising SKIP lines — e.g. a font that can't be found on disk — first.

2. Fix with visual sampling:

   unsubsetter --fix --verify-visual 10 interior.pdf
   

   This writes interior.unsubset.pdf and pixel-diffs 10 random pages against the original.

3. Independent structural check:

   pdffonts interior.unsubset.pdf
   

   Confirm sub=no on every previously-subset CID TrueType or CID CFF font.

4. Spot-check a few pages in a PDF viewer, paying attention to pages that use fonts the tool reported as skipped — those pass through unchanged and should look identical.

5. Upload to KDP. If it flags a different font, re-run with --only THAT_FONT to test it in isolation, or report the issue.

Troubleshooting exit code 4

If unsubsetter --fix exits with code 4, the disk font on your system doesn't match the subset embedded in the PDF for one or more CFF fonts. The report names the offending fonts. Either:

• Locate the matching font version and supply it via --font-path /path/to/font/dir; or
• Re-run with --exclude FONT_NAME to leave that font alone (it'll stay subset in the output).

Limitations

V2 handles CID TrueType (CIDFontType2) and CID CFF (CIDFontType0) subsetted fonts. Simple Type 1 (/FontFile) and Type 1C (/FontFile3 /Subtype Type1C) are detected and reported, but left unchanged. If a preflight checker flags one of those, outlining (converting the affected glyphs to vector paths) is the usual workaround until those types are supported.

For CFF fonts, unsubsetter runs a glyph-correspondence check between the embedded subset and the disk full font. If they don't agree on glyph identity (e.g., your disk font is a different version than the one originally embedded), the run exits with code 4 and writes no output. Either supply the matching font via --font-path or skip that font with --exclude.

Development

Set up the project and run the test suite:

uv sync --extra dev
uv run pytest