WOFF2 to TTF Converter

What WOFF2 does to a font

WOFF 2.0 packages an sfnt font for delivery, and it does two things WOFF 1.0 does not. It compresses the whole font with Brotli rather than compressing each table separately with zlib, and it applies a reversible preprocessing transform to the glyf, loca and hmtx tables before compression. The transform re-encodes point coordinates, flags and bounding boxes into separate streams that the entropy coder models far better than the raw table does. The W3C describes the goal as "improved compression and thus lower use of network bandwidth, while still allowing fast decompression even on mobile devices".

Every one of those steps is defined to run backwards. Decoding reconstructs the original tables, so unwrapping a WOFF2 gives you the same font data that was packaged, not a re-rendered approximation of it.

The glyf transform only applies to TrueType outlines. A CFF-flavoured font is stored with the null transform and simply Brotli-compressed, which means decoding a PostScript-outline web font hands you an OTF. That is the usual reason a "WOFF2 to TTF" result does not look like a TTF: nothing went wrong, the source was never TrueType. Converting it properly is an OTF to TTF job, and unlike this step it is lossy.

What you get back is the subset, not the family

One thing no decoder can undo is subsetting. Web fonts are usually cut down before they are packaged: the build strips glyphs the site never renders, often drops GSUB and GPOS layout features it does not use, and sometimes removes hinting to save bytes. Decoding the web font returns exactly that subset. If the file you pulled off a site carries 240 glyphs, the family it came from may have had two thousand, and the missing ones are not recoverable from the file you have.

WOFF2 also has a collection directory, so a single .woff2 can wrap several faces that share tables, the same arrangement a .ttc uses on the desktop. Decoding one of those yields a collection rather than a single font, and tools that assume one face per file will take only the first.

When you need the TTF at all

• Installing the face locally so a mockup matches what the browser renders.
• Opening it in a font editor or inspector with no WOFF2 support.
• Feeding a renderer, PDF library or build step that only reads sfnt files.
• Reading the name, head and OS/2 tables to identify the font and its embedding permissions.

If the goal is the opposite direction, packaging a desktop font for the web, start at TTF to WOFF2 instead. Round-tripping a font through WOFF2 and back gains nothing.

Related font conversions

• WOFF to TTF for the older zlib-based container.
• TTF to WOFF2 to package a font for delivery.
• OTF to TTF when the decoded file is CFF-flavoured.
• Font to Base64 to inline a small face into a stylesheet.
• Format hubs: WOFF2, TTF, WOFF and OTF.