Remove Accents/Diacritics

About the Remove Accents/Diacritics Tool

Most accented letters (é, ñ, ü, â and similar) are Unicode characters built from a plain base letter plus a separate "combining mark" glued on top. This tool first runs your text through normalize("NFD") to split each accented letter back into that base letter and its mark, then deletes every combining mark in one pass. A handful of letters don't work this way because they are not a base letter plus a mark in Unicode, ß, æ, œ, ø, đ, ł, ð and þ (plus their capital forms) are their own standalone characters, so those are swapped using a small hand-written lookup table (ß to ss, æ to ae, œ to oe, ø to o, and so on) before the normalization step runs.

• NFD normalization, decomposes precomposed accented letters into base letter + combining mark, then strips the marks.
• Fallback map, handles the letters that Unicode doesn't decompose, like ß, æ, œ, ø, đ, ł, ð, þ.
• Characters changed, counts how many original characters were altered or replaced by the conversion.

NFD specifically, not NFKD — canonical decomposition only

Unicode defines four normalization forms, and this tool uses NFD (canonical decomposition without recomposition), not NFKD (compatibility decomposition). The difference matters at the edges: NFD splits an accented letter into its base letter plus a combining mark, which is exactly what's needed here, but it leaves compatibility characters alone — things like the "fi" ligature (a single Unicode character, not two letters) or full-width Latin letters sometimes used in East Asian typesetting. NFKD would additionally unpack those into their ordinary equivalents, but this tool deliberately doesn't reach that far, since the goal is stripping accent marks specifically, not normalizing every visually-equivalent character variant in Unicode.

The combining-mark range covers the common case, not every diacritic block

After NFD decomposition, any character whose code point falls between U+0300 and U+036F is dropped — this is the "Combining Diacritical Marks" block, which covers the accents used on virtually all everyday Latin-script text: acute, grave, circumflex, tilde, umlaut, cedilla, and the rest of the marks that show up on café, naïve, or Zürich. Unicode does define a few other, much rarer combining-mark blocks outside that range, used mostly for specialized phonetic transcription or historical text, and those wouldn't be caught by this specific range check. For essentially all ordinary accented text in Latin-alphabet languages, though, this single block covers everything you'll actually encounter.

Every loop iterates by Unicode character, not by raw string index

Both passes through the text use for...of, and the total character count uses Array.from(), rather than a plain indexed loop or .length. That distinction matters because for...of and Array.from() both iterate by full Unicode code point, correctly treating a multi-byte character like an emoji as a single unit, where a naive index-based loop or raw .length call would split it into two meaningless surrogate halves. This is a more careful approach than this site's own Line Counter, whose character-length stat uses plain .length and so does inflate slightly on text containing emoji — a difference worth knowing about if you're comparing character counts between the two tools on the same input.

The fallback map runs before NFD, because those letters don't decompose either way

The order here is fallback substitution first, NFD decomposition second. It works this way because letters like ß, æ, ø, and the rest of the fallback list aren't a base letter plus a combining mark in Unicode at all — they're their own standalone characters, so running NFD on them wouldn't change them regardless of when it happens. Substituting them first just means the generic NFD-and-strip pass afterward only ever has to deal with genuinely decomposable accented letters, keeping the two steps cleanly separated by what kind of character each one is built to handle.

"Characters Changed" counts affected input characters, not the net change in length

The counter increments once for every original character that gets altered, whether that character is deleted (a combining mark) or replaced (a fallback substitution) — but several fallback substitutions actually expand into more than one output character. ß becomes "ss," æ becomes "ae," and þ becomes "th," each counted as a single "changed" character even though the output is two characters long for each one. So the stat measures how much of your original input was touched, not the difference in total length between the input and output text, which can end up slightly longer than what went in in text with several of these expanding letters.