Stylised game fonts are Unicode, not fonts

What you copied is not a font

Search for an "Adopt Me font" or a "Roblox font" and you land on a page of boxes with copy buttons. Whatever comes out of those boxes is not a typeface that gets installed anywhere. Unicode encodes characters, not glyphs: how a letter is drawn belongs to the font, but which letter it is lives in the code point stored in the file. Those decorative letters are not styled versions of the alphabet. They are different characters.

The practical consequence is simple. You cannot install a font into someone else's game. Adopt Me and Roblox render their interface with their own typeface and never look at what is installed on your machine. What you type into a name field travels to the server as a sequence of characters, so the only available mechanism for "changing the font" in a name is substituting characters that happen to look like letters. The Unicode text styles tool does exactly that, and it does it in the browser from a lookup table, so the text you type is never sent anywhere.

Which blocks these come from

Almost every style you will meet comes out of five blocks. Knowing the block is worth the effort, because the block largely determines where a style survives and where it falls apart.

Underlined, struck-through and zalgo styles work differently. They keep the ordinary letter and append a combining mark from U+0300–U+036F, so the letter really is the letter with something stacked on it. That is why an underlined name doubles in character count. If you want to inspect an individual code point, the Unicode symbol browser lists the block and encoding for each one.

Real font or Unicode trick

Four quick tests settle it every time.

1. Plain text test. Paste into the address bar or a text editor. If the styling survives it is Unicode. If it vanishes it was formatting, and formatting cannot travel into a name field at all.
2. Search test. Use Ctrl+F on the page with ordinary letters. If the styled text is not found, the characters genuinely differ from the ones you typed.
3. Length test. Every letter from the mathematical alphabets costs four bytes in UTF-8 against one for a plain letter. In fields with a character budget, a three-letter flourish can be enough to hit the ceiling.
4. Origin test. A copy button means Unicode. A .ttf or .otf download means a real font, and a real font only helps inside graphics you produce yourself.

Characters that quietly disappear

None of these blocks is a complete alphabet. Where a character is missing the converter passes it through untouched, so the name comes out half styled and nothing tells you why.

• Accented and non-English letters. Characters such as ç, ğ, ö, ş and ü have no styled counterpart in any of these blocks, so a name mixing them with ASCII renders inconsistently.
• Digits. The squared, filled-circle and parenthesized alphabets have no standard set for 0-9, and the italic alphabets have no digits encoded at all. In those sets digits pass through plain, which anyone whose name carries a number sees immediately.
• Parenthesized letters. Only lowercase ⒜ to ⒵ exists. Uppercase reuses the same glyphs, so capitalisation is lost entirely.
• Small capitals. There is no small-cap x, so "Max" comes out as "ᴍᴀx" with one letter standing a head taller than the rest.
• Subscripts. The letters b, c, d, f, g, q, w, y and z have no subscript form, which leaves most names roughly half converted.

The rule is easier to read backwards: the more exotic a style looks, the more gaps its alphabet has. The bold sans set covers all 26 uppercase, 26 lowercase and 10 digits. The italic sets carry no digits of their own, because Unicode never encoded italic digits at all, and the decorative sets have far wider gaps.

Where it works and where it gets rejected

The same string does not behave the same way in every field. Username fields are usually validated against a narrow allowlist, while display name fields tend to be permissive. When a style is rejected, the problem is the field's validation rule rather than the style itself.

To see which styles survive on a given platform, the sets on the Adopt Me name studio, Roblox text styles, Discord and Instagram are already trimmed to that field's character budget. Even so, the only reliable check is trying to save the name for real.

The accessibility cost nobody mentions

The official character names decide this one. U+1D5D4 is called MATHEMATICAL SANS-SERIF BOLD CAPITAL A. It is not a bold A. It is a separate symbol with its own identity, reserved for mathematical notation.

Screen readers speak text according to character identity. When a word cannot be recognised, the reader falls back to announcing characters one by one by name, or skips them. A name written in a mathematical alphabet therefore arrives as a run of symbol names, as repeated "unknown character" announcements or as nothing at all. A player using assistive technology cannot learn who you are. Combining-mark styles are worse still, because an extra mark follows every letter and the reading is shredded.

This is not a cosmetic complaint. The WCAG readability guideline expects text to be understandable to the machinery that presents it, and swapping letters for symbols attacks precisely that layer. The same substitution puts the name in the wrong place when a list is sorted, keeps it out of search results and removes it from autocomplete and mention lookups.

It also explains why platforms filter these characters in the first place. Profanity and harassment filters match ordinary letters, and the same word written in a mathematical alphabet slips straight past them. Many games and social networks therefore block the blocks outright in username fields.

Normalisation can undo all of it

NFKC, one of Unicode's normalisation forms, applies compatibility mapping: it folds mathematical alphabets, fullwidth forms and circled letters back to plain ASCII. On a system that normalises input that way, 𝗔𝗕𝗖 is silently stored as ABC. The save succeeds, the styling is gone and nothing explains where it went. When a name works on one service and comes out plain on another, this is usually the reason.

The honest split: fonts for graphics, Unicode for names

Treating these as two separate jobs makes the decision easy. For thumbnails, banners and channel art you use a genuine font: load a .ttf or .otf you are licensed for into a graphics program and you get the weight and kerning you actually wanted. Nothing about Unicode matters there, because the output is a picture.

For names, bios and chat, Unicode is the only mechanism available, so the work goes into choosing well. Pick a set with a complete alphabet, check that your digits and accented letters survived before you copy, and keep the decoration down to a character or two. For separators and frames an emoji is usually the safer choice, because emoji are rarely filtered and they carry a real spoken name for assistive technology. The emoji picker copies the character along with its code point.

Short checklist

1. Paste the finished name into a plain text field and confirm every letter converted.
2. Choose the bold sans set if your name has digits; accented letters convert in no set at all.
3. Aim at the display name rather than the username; sign-up fields are far stricter.
4. Keep combining-mark styles out of names, and out of anything longer than a short status line.
5. For graphics, use a licensed font file rather than Unicode substitutes.