Stripping accents from text online means converting letters like é, ñ, ü, ø, ł and đ into their plain Latin equivalents — café becomes cafe, Zürich becomes Zurich, Łódź becomes Lodz — so the result is usable in URLs, filenames, ASCII-only databases and contact lists. The fast route most people take is a one-liner that normalizes the string and drops the combining marks, but that route quietly fails on an entire family of letters (ø, ł, đ, ħ, þ and the ligatures æ, œ, ß) because Unicode records no decomposition for them, so nothing pops off and the character passes through unchanged. A browser-based tool that pairs that approach with an explicit fold table — and that refuses to touch Greek, Cyrillic, CJK or emoji in the process — is what the search "remove accents from text online" actually wants. Remove Accents from Text does exactly that: it strips true diacritics from Latin base letters, folds the non-decomposing letters through a table verified against the Unicode Character Database, and reports the exact number of characters changed.
Why a normalize-and-strip script quietly fails
The method that almost every quick online script reaches for is two lines long. You call String.prototype.normalize("NFD") on the input, then delete every Unicode character whose category begins with "M" — the combining marks. For most accented Latin text that does the right thing: é splits into e + combining acute, the acute gets deleted, and the output is plain e. The Unicode Consortium documents this behavior in UAX #15, the report that defines the NFD, NFC, NFKD and NFKC normalization forms, and the JavaScript engine implements it exactly as specified. So far, so good.
The failure shows up the moment the input contains letters whose "accent" is not actually a separate mark. Unicode records no decomposition mapping for letters such as ø, Ø, đ, Đ, ł, Ł, ħ, Ħ, ŧ, Ŧ, ı, ð and Ð, which means NFD leaves them as single code points and your deletion loop has nothing to delete. Łódź goes in as Łódź and comes out as Łódź. The same problem hits the Nordic and Latin ligatures: æ, œ, ß, þ and their capitals are single code points whose Unicode decomposition is empty, so they pass through any naive pipeline without changing.
This is not a bug in NFD; it is exactly the behavior the standard describes. But it means that for a Polish city name, a Danish surname or a German street name, the common one-liner is silently wrong. The Remove Accents from Text tool closes that gap by carrying an explicit fold table for those letters, with every entry checked against the Unicode Character Database, so the same Polish, Danish or German text becomes Lodz, Odense and Strasse on the first pass instead of surviving normalization unchanged.
How to remove accents from text online
- Paste the text that contains accented or special Latin characters into the input box. Up to one million characters can be processed in a single pass, so a contact list, a CSV export or a long article body all fit without splitting or batching.
- Choose how the edge cases should be handled. One toggle controls ligatures and letters that need to be spelled out (æ to ae, œ to oe, ß to ss, þ to th, plus their capitals). It is on by default because data cleaning almost always wants those folds, and you can switch it off to keep the original letters while still stripping true diacritics. A second toggle controls whether non-Latin scripts should be kept (the default) or stripped entirely when you truly need ASCII-only output.
- Click Remove accents and watch the change counter. The page reports exactly how many characters were altered, which lets you confirm that the input really was touched and gives a rough sense of where the modifications happened. The cleaned result is then one click away in the output box, ready to copy into a slug, a database or a code file.
The operation is idempotent, which means feeding the cleaned output back through the tool changes nothing further. Precomposed input (é as one code point) and decomposed input (e followed by a combining acute) both produce the same final string, because the implementation applies NFD, removes combining marks from Latin base letters only, runs the fold table, and recomposes to NFC before returning the result.
What gets changed and what stays
The table below shows how a handful of common inputs behave under three different policies: a naive normalize-and-strip script, the Remove Accents from Text tool with both toggles at their defaults, and the same tool with the transliteration toggle switched off.
| Input | Naive NFD only | Remove Accents from Text (default) | With transliteration off |
|---|---|---|---|
| café | cafe | cafe | cafe |
| Zürich | Zurich | Zurich | Zurich |
| Łódź | Łódź (no change) | Lodz | Lodz |
| naïve | naive | naive | naive |
| señor | senor | senor | senor |
| straße | straße (no change) | strasse | straße |
| Æsop | Æsop (no change) | Aesop | Æsop |
| Άλφα (Greek) | Αλφα (corrupted) | Άλφα (untouched) | Άλφα (untouched) |
| Привет (Cyrillic) | Привет (corrupted) | Привет (untouched) | Привет (untouched) |
The transliteration toggle is split out on purpose. Turning ß into ss, þ into th, æ into ae or œ into oe is not the same thing as removing an accent; it is spelling one letter as two. The tool labels that as transliteration and keeps it behind its own switch, instead of silently mixing the two operations, which matters whenever you need to round-trip text or preserve authorial spelling. The default behavior of both toggles matches what data-cleaning pipelines expect, but every step is visible and reversible from the same screen.
Where stripping accents actually pays off
The single most common reason people search "remove accents from text online" is slug generation. URLs, file names, blog post identifiers and Git branch names all want predictable ASCII, and a single stray é in café-recipes.md can break downloads on a misconfigured server or get percent-encoded in a way that hurts search rankings. Running the title through the tool first turns that path into cafe-recipes.md on the spot, with no regex writing required.
Contact deduplication is the second big win. Two copies of the same person in a CRM — "Łukasz Kowalski" and "Lukasz Kowalski", or "Stéphanie Müller" and "Stephanie Muller" — are treated as different records by an exact-match join. Once both sides are passed through the folder, the records collapse into one. The same logic applies to name matching across systems during a migration, building identifiers that have to be stable across encodings, and preparing CSV exports for software that does not speak UTF-8.
The third family of uses is around analysis. Word frequency, search indexing and text mining all normalize input before counting tokens, and stripping accents is part of that normalization. The tool's idempotence guarantee means it is safe to run inside a repeated pipeline: each pass is a no-op after the first, so it will not slowly eat letters out of your corpus, and the explicit change counter gives you a quick sanity check that the right number of substitutions actually happened.
Limits to know before you paste
This is a Latin-script diacritic remover with documented transliterations, not a general transliterator. It will not romanize Russian Cyrillic into Latin letters, will not turn Greek into a phonetic spelling and will not convert Chinese characters into pinyin. The fold table covers the Latin-script letters that the Unicode Character Database explicitly omits from decomposition, plus the most common ligatures; specialist letters such as ĸ are folded by convention rather than by any Unicode rule, and the page states that openly instead of hiding it.
Greek and Cyrillic deserve a separate callout because they decompose in Unicode exactly the way accented Latin does: ά splits into α + combining acute, ё splits into е + combining diaeresis. A naive stripper quietly rewrites both to α and е and corrupts the script. Remove Accents from Text only removes marks attached to Latin base letters, so Greek, Cyrillic, Chinese, Japanese, Korean and Devanagari pass through byte-for-byte, and Devanagari vowel signs — which are grammatically letters in their own right rather than accents — are never treated as removable marks. Emoji behave the same way: they survive untouched, including multi-codepoint sequences like flag emoji and skin-tone modifiers.
Two operational details round out the contract. First, input is capped at one million characters and processes in a single linear pass, so even huge pastes return quickly. Second, everything runs in the browser — nothing is uploaded to a server, nothing is stored, and no account is required, which matters when the text you are cleaning is itself sensitive, such as a customer roster or a private document.
If you're weighing options, How to Remove Duplicate Lines in Word Documents covers this in detail.
If you're weighing options, How to Remove Empty Lines in a Word Document covers this in detail.