Why the Slug Generator Doesn't Turn café into caf-zrich

Guide · Last verified Aug 26, 2026

Say you're turning the blog title "Café Zürich" into a slug. The simplest possible implementation is a single regex: "strip anything that isn't a letter or digit." Do that, though, and é and ü vanish entirely, leaving the nonsensical result "caf-zrich." The original letters were just deleted — not converted to "e" or "u." The proper way to solve this is Unicode NFD normalization, and the Slug Generator actually uses exactly this approach.

1. The root cause: an accented character is a "different" character

A person reads é as "e with an accent on it," but to a computer, é is fundamentally its own, completely separate code point (U+00E9). A regex like [^a-z0-9] filters it out because é doesn't fall in the a-z range — it just gets deleted, not replaced with "e." Strip é out of café and you're left with "caf"; strip ü out of Zürich and you're left with "zrich," producing the unreadable slug "caf-zrich."

2. Unicode normalization: splitting one character into two pieces

Unicode has two ways to represent the same letter. One stores é as a single code point (U+00E9, the composed form, NFC). The other stores it as two separate code points: e (U+0065) plus a combining acute accent mark (U+0301) — this is NFD, Normalization Form Decomposition. Both look identical as é on screen, but the underlying data is structured completely differently. JavaScript's String.prototype.normalize('NFD') decomposes a string into that second form: base letter plus combining mark.

3. A regex that removes only the combining marks

Unicode groups combining diacritical marks together in a dedicated block, code range U+0300 through U+036F. After decomposing with NFD, removing only characters in that range with a regex leaves the original base letters (e, u, n, c, and so on) untouched and strips only the accent marks. The result is the natural conversion people expect: é → e, ü → u, ñ → n. Unlike blindly deleting non-ASCII characters, this approach "strips the decoration" rather than "deleting the letter."

4. What the slug generator's code actually does

Inside this tool's generateSlug() function is the line v.normalize('NFD').replace(/[̀-ͯ]/g,''). This decomposes the string with NFD, then uses a regex to remove only characters in the U+0300–U+036F range (combining diacritical marks). This step runs before the "remove special characters" option, so text like café or Zürich doesn't lose its letters entirely — it naturally becomes cafe and zurich. After that, if "remove special characters" is enabled, anything besides letters, digits, whitespace, hyphens, and underscores is stripped, and finally consecutive runs of whitespace and separator characters are collapsed into whichever separator you chose. As the codebase confirms, the actual implementation does use .normalize('NFD').

5. Why non-Latin scripts like Korean and Chinese are handled differently

NFD normalization only helps with Latin-family accented characters that have a "base letter + diacritic" structure. Korean (Hangul) and Chinese characters were never built from a separable base-letter-plus-accent structure, so NFD doesn't decompose them at all — they simply fall outside the letter/digit range at the "remove special characters" step and get deleted as-is. So if you need an SEO slug from a Korean title, it's recommended to translate or transliterate it into English first and then run it through this tool. If you also want to polish the title itself before turning it into a slug, the SEO Title Generator is a useful companion.

Frequently Asked Questions