@0x6b/ttf2woff2-wasm

ttf2woff2-wasm

Convert TTF fonts to WOFF2 format. WASM build of 0x6b/ttf2woff2.

Drop-in replacement for nfroidure/ttf2woff2 with zero native dependencies.

[!NOTE] This package uses a different Brotli implementation, so the output may differ in size from nfroidure/ttf2woff2. Please verify the output fonts work correctly in your target environment before deploying to production.

Installation

$ npm install @0x6b/ttf2woff2-wasm

Usage

const ttf2woff2 = require("@0x6b/ttf2woff2-wasm");
const fs = require("fs");

const ttf = fs.readFileSync("font.ttf");
const woff2 = ttf2woff2(ttf);
fs.writeFileSync("font.woff2", woff2);

Migration from nfroidure/ttf2woff2

package.json:

-"ttf2woff2": "^3.0.0"
+"@0x6b/ttf2woff2-wasm": "^0.14.0"

Your code:

-const ttf2woff2 = require('ttf2woff2');
+const ttf2woff2 = require('@0x6b/ttf2woff2-wasm');

const woff2 = ttf2woff2(ttfBuffer);

API

ttf2woff2(input: Buffer | Uint8Array, options?: { transformGlyfLoca?: boolean }): Buffer

Convert TTF font data to WOFF2 format synchronously.

Encoding behavior

The WASM encoder uses Brotli quality 11. By default, transformGlyfLoca is false: the original glyf/loca tables are preserved without transformation and are still Brotli-compressed. Disabling transformation avoids glyph reconstruction that can make some fonts incompatible with Chromium and Google's WOFF2 decoder, at the cost of potentially larger output. Existing one-argument calls use this default.

To explicitly enable transformation:

const woff2 = ttf2woff2(ttf, { transformGlyfLoca: true });

Transformation can reduce output size, but reconstructed glyph offsets can exceed the range of a short loca table, causing affected fonts to be rejected by Chromium or Google's decoder. Validate transformed output with your target browsers and decoders before using it.

Benchmark

Measured on Linux x64 in an Amp Orb (8 vCPUs, Intel Xeon @ 2.60 GHz), Node.js v26.10.0, using ttf2woff2 v8.0.0 and a release build of ttf2woff2-wasm v0.14.0. Results are specific to these fonts and this environment.

Conversion Performance

Each variant runs sequentially with 3 warm-up conversions followed by 10 timed conversions. Input files are read before timing, module initialization is excluded, and garbage collection is requested before each timed batch via --expose-gc. The script reports mean, minimum, maximum, and output size; the table shows the mean.

The native measurement loads the compiled addon directly and fails if it is unavailable, rather than silently measuring that package's WASM fallback. WASM uses Brotli quality 11: default leaves glyf/loca untransformed, while opt-in sets transformGlyfLoca: true. These modes have different size/compatibility tradeoffs; opt-in is not a recommendation for all fonts. See Encoding behavior.

Memory usage is not compared: JavaScript heap deltas do not capture native allocations, WASM linear memory, or peak process memory.

Run from the repository root:

$ cargo xtask build
$ cd bench
$ npm install ttf2woff2@8.0.0
$ node --expose-gc benchmark.cjs

Package Size

Sizes are the npm tarball contents after unpacking, excluding dependencies and generated native build artifacts, not total installation disk usage. The WASM size is from the local release build (npm pack --dry-run --json ./pkg); the native package size is from npm metadata (npm view ttf2woff2@8.0.0 dist.unpackedSize). The WASM package requires no native compilation.

License

• The Noto Sans Japanese font in tests/fixtures/ is licensed under OFL.
• The WarpnineSans font in tests/fixtures/ is licensed under OFL.
• Everything else is dual-licensed under MIT or Apache-2.0.