Nobody Told Me Unicode Was This Weird — So I Built a Font Generator

Chris Xhangrei — Wed, 17 Jun 2026 11:55:14 +0000

Someone in a Discord server changed their username to something that looked like a different alphabet, except it wasn't. I copied it into a random site, pasted my own name, and it came back looking like 𝔠𝔥𝔯𝔦𝔰. Sent it to a friend. He asked how I did that. I had no idea.

That should have been the end of it. I pasted the text, it worked, move on. But the question stayed: Discord doesn't support custom fonts. Nobody downloaded a plugin. The text just showed up different on everyone's screen. How?

The answer broke my brain a little. These are not fonts. 𝔠 is not the letter c. It looks like c. It has the same phonetic role in that word. But to the computer, it is a completely separate character — its own entry in a database of every symbol that computers are supposed to represent. The font generator doesn't load a different typeface. It finds a character that looks like the letter you typed and swaps it in. Your device's default font renders that character. That's it. The "font" was always your own system.

I had to sit with that for a minute.

Unicode is the system that assigns every character a code point — a number that tells the computer what to display. There are over 149,000 characters in the standard. One organization, the Unicode Consortium, maintains the whole thing. They review proposals, vote on additions, publish a new version roughly every year. Emoji go through this same process. The face-with-tears-of-joy emoji has an official code point, a proposal document, and a ratification date. So does the cuneiform sign for ox.

Where the "font styles" actually come from is the weird part. That bold text people use on Instagram and LinkedIn — 𝐁𝐨𝐥𝐝 𝐥𝐢𝐤𝐞 𝐭𝐡𝐢𝐬 — those characters live in a block called Mathematical Alphanumeric Symbols. They were added so that math papers could distinguish between variable types in digital documents. Researchers needed a way to write a vector 𝒗 differently from a scalar v in the same document, without relying on a renderer to apply formatting. So they got their own code points. The internet found those code points and decided they looked cool for bios.

The rune situation is even better. Elder Futhark runes are in Unicode. The proposal came from historical linguists doing serious work on early Germanic inscriptions. The Consortium approved them. They are now used to write PUBG usernames. Nobody in those proposal meetings saw that coming.

There's a whole separate problem with characters that don't show up. Unicode has 149,000+ entries, but your device's font only has so many of them mapped. If a character exists in the standard but your font doesn't have a glyph for it, you get a box. Or nothing. What renders on iOS is not guaranteed to render on Android, and neither is guaranteed to work in a specific game client. Discord has broader coverage than WhatsApp. A PUBG client has whatever the developer bundled. No published list exists of what works where. Players figure it out by pasting things in and checking, which is honestly the only way to know.

I started building the tool because I couldn't find one that didn't look like it was made in 2008. Static site, no backend, all the conversion runs in the browser. 168 styles total. I thought mapping the character ranges would take an afternoon. It did not. Some ranges have gaps — code points that exist but don't have a clean visual equivalent, so a character drops out mid-word, which looks broken. Other ranges work fine in isolation but behave strangely when mixed. I spent more time on edge cases than on anything else in the build.

If you want to see what 168 Unicode styles looks like in one place, it's at aifontsgenerator.com.

The thing I keep circling back to is the Consortium itself. These are people who built a standard for encoding the world's languages — historical scripts, mathematical notation, scientific symbols. Serious work, dense documentation. And somewhere in there, the internet turned their character database into an aesthetic toolkit. Zalgo text — that corrupted glitchy horror thing from creepypasta — uses combining characters that were added for languages where diacritical marks stack above and below letters. A legitimate linguistic feature, used for stacking fifty marks on a single e until it looks haunted.

That's a real sentence. I'm still not sure what to do with it.

DEV Community: Chris Xhangrei

Nobody Told Me Unicode Was This Weird — So I Built a Font Generator