How to use
- Add font files (TTF, OTF, WOFF, WOFF2). You can add several at once.
- Pick the output format: WOFF2 for the web, TTF/OTF to install on a desktop.
- To shrink the file further, turn on Subset and choose the characters to keep (presets, pasted text, Unicode ranges).
- Press Convert. A table shows original and result sizes and how many characters were kept. Download files one by one or as a ZIP, and copy the @font-face code into your CSS.
Output formats
| Format | Notes |
|---|---|
| WOFF2 | The web standard. Glyph data is transformed and then Brotli-compressed, so it is the smallest. Every modern browser supports it. |
| WOFF | A web font with each table zlib-compressed. Use it only when you need very old browsers. |
| TTF/OTF | Uncompressed, installable. Use it to unpack WOFF or WOFF2. CFF-outline fonts are saved as .otf. |
Subsetting: keep only the characters you need
Korean fonts are large because Hangul alone has 11,172 syllables. Subsetting drops every glyph except the ones you choose, and keeps only the features those characters need, such as kerning (kern) and ligatures (liga).
| Characters kept | WOFF2 size from this tool |
|---|---|
| The whole font | 2,057,544 bytes |
| Hangul 11,172 + basic Latin | 1,718,552 bytes (about 16.5% smaller) |
| Hangul 2,350 (KS X 1001) + basic Latin | 426,808 bytes (about 79.3% smaller) |
| 251 digits, symbols and punctuation | 27,548 bytes |
| 95 basic Latin characters | 23,216 bytes (about 98.9% smaller) |
These were measured on Pretendard Variable 1.309, a variable font that is 6,759,776 bytes as TTF. The same font saved as WOFF (zlib) is 2,982,860 bytes and as WOFF2 is 2,057,544 bytes, about 55.9% and 69.6% smaller than the TTF. Subsetting keeps the variable wght axis intact. Results differ per font, so read the real numbers from the Characters kept and size columns.
Be careful: characters missing from a subset are drawn by the browser with a fallback font. Where users type their own text (comments, search boxes), use the full 11,172 syllables or the whole font.
Splitting with unicode-range
If you split a subset into several files and give each @font-face a unicode-range, the browser downloads only the files whose characters the page actually uses. The code this tool generates records the characters that really remain in the subset as unicode-range, merging consecutive runs (U+20-7E). Hangul 2,350 plus Latin comes to 1,614 ranges (about 14 KB).
@font-face {
font-family: "Pretendard";
src: url("Pretendard.subset.woff2") format("woff2");
font-weight: 400;
font-display: swap;
unicode-range: U+20-7E,U+AC00-D7A3;
}
font-display: swap shows text in a fallback font while the web font loads, which shortens the time text looks missing.
Subset options
| Option | What it does |
|---|---|
| Keep font features | Keeps common features such as kern and liga. Turning it off drops all feature data for a smaller file, but kerning and ligatures disappear. |
| Remove hinting | Drops the instructions that fine-tune shapes at small sizes. Fonts with almost no hinting, such as variable fonts, barely change. |
| Keep the name table | Keeps copyright and license text in the result. Leave it on. |
Check the license
Every font has its own terms for web conversion, subsetting and redistribution. Before converting, this tool shows the license text stored in the font and its embedding restriction (fsType), but it cannot judge what is legally allowed. Fonts under the SIL Open Font License (OFL) can usually be subset, though there may be separate conditions on using the font name. For commercial fonts, check the web license separately.
Related tools
To see which characters a font contains before converting, use Font Viewer. To bundle the output, use ZIP Creator; to confirm what a file really is, use File Type Checker; to tidy the CSS you copy, use CSS Beautifier & Minifier.
Your files are processed only in your browser and are never sent to a server.