Character counter

What this character counter calls a character

The large character total counts grapheme clusters: the symbols a person normally experiences as single characters. An accented letter typed as e followed by a combining accent is one grapheme. A family emoji joined from several people and invisible joiners is one grapheme too. Counting JavaScript string positions would split both cases, so that number appears separately as UTF-16 code units rather than being passed off as the answer.

Spaces are included in the main total because they occupy real positions in text. The receipt also shows a total without Unicode whitespace, which excludes ordinary spaces, non-breaking spaces, tabs and line breaks. It does not remove punctuation or invisible formatting characters. If the job is to find and name those characters, use the invisible character detector.

Code points, UTF-8 bytes and UTF-16 units are different answers

Unicode assigns a code point to each encoded element, but several code points can cooperate to draw one grapheme. UTF-8 then stores those code points in between one and four bytes. JavaScript stores strings as UTF-16 code units, where a code point outside the Basic Multilingual Plane takes a surrogate pair. None of these counts is more “correct” in isolation; each answers a different question, which is why the receipt keeps all of them visible and named.

Use graphemes when the limit is about what a reader sees, code points when inspecting Unicode data, UTF-8 bytes for an encoded payload, and UTF-16 units when an API defines length using JavaScript or another UTF-16 environment. The longer explanation in UTF-8 versus ASCII shows why even ordinary accented text can make byte count and character count diverge.

How words and line endings are counted

Words use the browser's Unicode word-boundary implementation rather than splitting only on ASCII spaces. Contractions such as don't stay together, and scripts that do not put spaces between every word can still be segmented. Word boundaries are linguistic, so this total is useful for document length rather than a grammar judgment. To rank repeated terms and export their shares, use the separate word frequency counter.

CRLF is one line boundary, not two. A pasted line feed and a decoded Windows line ending therefore produce the same number of lines. Opening a file preserves the decoded text for this analysis, while the UTF-8 byte total deliberately describes that text after decoding — it does not claim the source file was UTF-8.

Why there are no social-platform limits here

A platform limit is an algorithm and a product policy, not a universal character definition. X applies weighted counting to some Unicode and treats URLs specially; account tiers can have different limits. SMS chooses GSM-7 or UCS-2 and charges by segments, so a single unsupported character can change the capacity of the whole message. Search titles are truncated to fit a rendered width rather than a published character ceiling. A generic counter that paints one of those numbers green would be precise-looking and wrong.

Everything here runs locally. Short text and multi-megabyte files use the same pure count, but the work runs outside the page's main thread so Unicode segmentation does not freeze typing. Choosing another file or editing the text cancels the stale run; only the receipt for the current text is allowed to appear.