Font Subsetting

dodeca scans all rendered HTML to find which characters are actually used, then strips everything else from your font files. A 300KB font typically becomes 15-30KB.

After rendering, we collect every character that appears in the output. For each font file, fontcull removes unused glyphs and OpenType features. The subsetted font gets a content hash and CSS @font-face rules are rewritten to match.

Most edits don't introduce new characters, so the subsetted font stays cached. When the character set does change, only fonts containing those characters are re-subset.

Variable fonts work too—subsetting preserves variation axes while removing unused glyphs.

fonts/inter.woff2: 285KB → 24KB (91% reduction)
fonts/fira-code.woff2: 178KB → 18KB (90% reduction)