Extract Unicode Graphemes

What a user perceives as a single character, even when it spans multiple Unicode code points. The ZWJ family emoji 👨‍👩‍👧‍👦 is one grapheme to the user but 7 code points to the encoder. The Greek flag 🇬🇷 is one grapheme but 2 regional indicators. A pre-composed é (U+00E9) is one code point and one grapheme; a decomposed é (e + U+0301) is two code points but still one grapheme. The Unicode Standard Annex #29 defines exactly when characters cluster.

Because a grapheme made of multiple code points is exactly what makes this tool interesting. Showing only the first code point of a family emoji misleads you about the underlying data – the size is hugely different (7 code points vs 1), and operations like string.length in JavaScript return code-unit counts that don’t match what you’d expect. The full sequence reveals the truth.

Three different ways to count: a code point is the abstract Unicode character ID (e.g. U+1F468 is “man”). A UTF-16 code unit is JavaScript’s internal storage unit – characters above U+FFFF take two code units (a surrogate pair). A UTF-8 byte is what gets written to disk or sent over the wire – ASCII is 1 byte, accented Latin is 2, most other BMP characters are 3, and astral characters (emoji) are 4. A single family emoji can be 1 grapheme, 7 code points, 11 UTF-16 units, and 25 UTF-8 bytes.

The stats line shows “engine: Intl.Segmenter” when the native API is available (Chrome 87+, Edge 87+, Safari 14.1+, Firefox 125+), or “engine: fallback (code-point split)” when it isn’t. The fallback iterates by code point – it handles surrogate pairs correctly but cannot recognise multi-code-point grapheme clusters like ZWJ sequences, so emoji families come out split. Practically, almost any browser from 2022 onward uses the proper API.

Yes. Intl.Segmenter is locale-aware and handles Latin, Cyrillic, Greek, Arabic, Hebrew, Devanagari, Chinese, Japanese, Korean, Thai, and all emoji. The tool initialises with locale 'en'; the segmentation rules for graphemes are locale-independent per UAX #29, so the locale doesn’t change the result for graphemes (it would matter for word/sentence granularity).

The native Intl.Segmenter is fast – millions of characters in under a second. The tool runs synchronously on the main thread, so very large inputs (10 MB+) might briefly lock up the page. The 150 ms input debounce keeps typing-while-editing smooth on any practical input size.

Cursor positioning, text selection, character counting (Twitter-style “200 characters”), and string truncation all need graphemes – not code units or code points – to behave the way users expect. Splitting a string on code units risks cutting a family emoji in half (broken render) or a surrogate pair in half (corrupt data). The CSV/JSON output of this tool is useful as a reference dataset when writing tests for grapheme-aware code.

No. The page loads three static files (HTML, CSS, JS) and then runs entirely in your browser. Your text never leaves the device – no fetch, no XHR, no analytics, no cookies. You can disconnect from the internet after the page loads and the tool still works.

Yes – free, unlimited, no signup, no watermark. The extracted graphemes are yours to use anywhere. Attribution to is appreciated but not required.