Accents, Diacritics & Fancy Fonts

When you pick a "bold" or "script" style, you're not applying formatting the way a word processor does. You're swapping each letter for a different Unicode character that happens to look bold or scripted — the bold 𝗮 is its own code point, not a normal "a" wearing a costume. (If that idea is new, start with why fancy fonts turn into boxes.)

Those styled characters live in a Unicode region called the Mathematical Alphanumeric Symbols block, and it was built for equations, not languages. It contains styled forms of exactly three things: A–Z, a–z, and 0–9. That's the entire alphabet available to a letter-swap style.

Every accented letter you can think of — á é í ñ ü ç, and every Vietnamese vowel like ữ ế ệ ơ ư ă — lives in completely different Unicode blocks (Latin-1 Supplement, Latin Extended-A/B/Additional). Those blocks have no styled twins. So when a generator reaches for a bold version of "é," there simply isn't one to reach for — which is why a naive tool leaves the é plain while its neighbours restyle, giving that half-styled look.

So the closed alphabet is real, but it isn't the end of the story. Unicode stores most accented letters as a base letter plus a separate combining mark, and NFD normalization pulls them apart. This generator does exactly that: it splits the accented letter, styles the base "n" like any other letter, then re-appends your original tilde — the same trick the strikethrough and underline styles always used for their own marks, now applied to accents across the board.

Feed an accented letter into a generator and one of three things happens to it.

Fate #1 is now the rule: Spanish, French and Vietnamese all style cleanly, tone marks included. Fate #2 is one family to avoid when accents matter — the upside-down group. (A second family, 13 novelty styles like the circled-bubble, fullwidth and katakana sets, also deliberately keeps the plain accented letter, because their exotic glyph blocks can't display a reattached mark cleanly — more on that in the table below.) And fate #3 is why a Vietnamese name with a đ in it can never be fully converted by any style, no matter what — the đ will always read as a plain d.

Here's the useful part. We ran every style in this generator through the engine — and then through a real browser, since combining marks depend on font fallback, not just valid Unicode. The result: 77 of 112 styles fully preserve accents, including every core typography style. The exceptions fall into a few named groups, each for a different reason.

The mechanism is the same one the strikethrough and underline styles always used for their own line: decompose the letter with NFD, act on the base, then re-append whatever combining marks were riding on it. A bold "ñ" doesn't exist in Unicode — but a bold "n" does, and the original tilde is a separate mark you can simply put back on top of it. Applied across the registry, that's how bold, italic, script and the rest went from stranding accents to keeping them.

Now that marks are reattached across most styles, this part matters more, not less — in fact it's exactly why the 13 exotic-block styles above keep the plain accent instead of a reattached one. Even when an accent is placed correctly, whether it looks right depends on the reader's device, browser, and screen — so you can't judge accented styled text from your own phone alone.

The browser flip: Chrome vs Firefox

When an accent is stored as a separate combining mark, Chrome, Edge and Safari quietly re-compose it onto its letter before drawing (they normalise to NFC). Firefox does not — it hands the sequence to the font as-is. So a reattached accent that looks perfect in Chrome can render as a floating, offset mark in Firefox on the very same page. Same text, opposite result, decided entirely by the reader's browser.

The platform gap: iOS vs Android vs Windows

• Windows can draw the styled glyphs at all only because a math font (Cambria Math) sits in its fallback chain — the default Segoe UI covers almost none of the block. In older or plain text controls with no fallback, styled letters become boxes.
• iOS is strict: if a character isn't in the system font, iOS often won't fall back to another font at all — the glyph simply shows as an empty box. Apple deliberately omits many of these code points.
• Android leans on its Noto fonts ("no more tofu") which cover the math block on modern versions — but Android 5-and-older and low-end builds never got that coverage, so the same styled name that's crisp on a new phone is tofu on an old one.

Size, zoom & screen density: the noticeability dial

Two independent things decide whether breakage is visible: whether the accent is placed correctly, and whether the screen has enough pixels to show it. They pull in opposite directions:

There's a counterintuitive lesson here: a Spanish name in a script font can look fine on a cramped mobile preview and visibly wrong when someone opens it full-size on a desktop. If you only check the small version, you're checking the case most likely to hide the problem.

With marks reattached, the accents themselves style in every Latin-script language. What changes from one to the next is only how many one-piece letters it uses — the atomic characters that still can't carry a mark. Rough guide, least to most affected:

So a Spanish or a Vietnamese name both style cleanly now, dấu and all — the only guaranteed holdout is đ. Which brings us to the language that stress-tests all of this.

Vietnamese is the stress test for every idea on this page, because a single vowel can stack a base letter, a vowel-shape mark (â, ă, ơ, ư) and a tone mark (sắc, huyền, hỏi, ngã, nặng). "ế" is really e + circumflex + acute — three code points. The old letter-swap behaviour restyled only the bare "e" and dropped the two marks that carry the meaning; the generator now decomposes the whole stack, styles the "e," and puts both marks back, so ế comes through styled and complete.

The NFC / NFD trap

Vietnamese text exists in two encodings that look identical: precomposed (NFC — "ệ" is one character) and combining (NFD — "ệ" is three). Most keyboards and websites use NFC, but text copied from a Mac often arrives in NFD. The styling engine handles both — it decomposes to NFD internally before reattaching marks — but the output then carries combining marks, so how cleanly it displays still depends on the reader's font and browser (the Chrome-vs-Firefox split above). The flip styles remain the exception: fed NFD Vietnamese, they can reorder a vowel and its tone mark independently.

The đ that never budges

As above: đ/Đ has no decomposition and no styled twin, so it's the one letter guaranteed to stay plain in every letter-swap style — and it also survives accent-stripping, so tools that "remove dấu" have to special-case it by hand.

What Vietnamese users actually do

Long before generators handled dấu, the Vietnamese scene had already routed around the problem, and the habits stuck. The popular category isn't "font" at all — it's "kí tự đặc biệt" (special characters): instead of restyling letters, users decorate a plain name with symbol frames (꧁ ༒ 亗 ★), which keeps the letters — and their dấu — safe. Restyling with the accents kept now works too, so it's an option rather than the only one. The common moves:

• Wrap, don't swap — put ornaments around a plain name; the accent-safe symbol styles above do exactly this.
• Bỏ dấu — deliberately drop the tone marks (Hưng → Hung) so the whole word styles uniformly. Standard for game names on Free Fire and Liên Quân, where names are length-capped and diacritics often error out.
• Test before committing — paste the candidate into a chat box first; if it shows □ or ?, pick another.

The lesson for anyone styling Vietnamese: a bold or script style now keeps your dấu, so restyle with the accents if you want them. Reach for bỏ dấu or symbol frames when a game's length cap or a platform's character rules force your hand — not because the font can't hold the marks anymore.

Even a perfectly styled accented string can change the moment it leaves the generator, because many systems normalise text on the way in.

• Search sees through the styling. Indexes and search boxes fold styled letters back to plain (NFKC normalization), so 𝐛𝐨𝐥𝐝 is indexed as "bold." That's good for being found — but it also means fancy text won't match a search for the normal spelling, and Ctrl-F / in-app search skips right over it.
• Some fields strip accents outright. A common sanitising step (decompose, then drop combining marks) erases every accent — and, again, leaves đ behind. Your carefully-accented text can arrive plain.
• Handles are ASCII-only everywhere. Instagram, Discord, TikTok, X, Telegram, Snapchat, Roblox and Steam all restrict the @handle to plain letters/numbers/dot/underscore. Styled and accented text belongs in the display-name field, never the handle.
• Stacking marks has limits. Platforms cap combining marks to fight zalgo (Discord stops handling them past ~150; Unicode's own "stream-safe" limit is 30). Piling zalgo on top of Vietnamese is doubly destructive — the glitch marks drown the tone marks that carry meaning.

And length counts against you: on X, each styled letter counts as two characters, so a styled accented bio hits the limit roughly twice as fast — while also raising the odds that at least one character lands on a device that can't draw it.

1. Assume your accents will style — then check the exceptions

Every core typography style now keeps accents, in any language, Vietnamese included. What to watch for: the upside-down styles (they don't flip accented letters), the 13 novelty exotic-alphabet styles (bubble circles, fullwidth, katakana and the like, which keep the plain accent), and the one-piece letters (đ ø ß ł ı) that style as a bare base shape.

2. Need a guaranteed-flat mark? A mark or wrap style is the safe fallback

Strikethrough, underline, slash and symbol-wrap styles preserve diacritics without decomposing anything — handy if a letter-swap style hits one of the exceptions above, or if you want to avoid reattached combining marks on a shaky rendering target.

3. Remember đ always drops to a plain d

đ has no accent to lift off, so in every style it converts to a bare styled "d" (or stays "d"). If a name leans on that đ to read correctly, keep the word plain, or accept the d — no style can produce a styled đ.

4. Style a word, not a paragraph

Every styled character costs on the accessibility side (a screen reader reads "mathematical bold small e" one glyph at a time) and adds another chance to break. Style a name or a hook, not a whole bio.

5. Test on the worst case

For accents specifically, that's an older Android phone and Firefox — the two renderers most likely to expose a dropped or detached mark. If it survives there, it'll survive nearly anywhere.