Character Counter

How to use the character counter

Paste or type into the box and four separate counts appear at once: graphemes, code points, UTF-16 code units, and UTF-8 bytes, plus the same text with all whitespace stripped out. On plain English text all four agree, and the tool is dull in exactly the way you want.

The moment an emoji, an accented letter, a non-Latin script, or a combining mark enters the string the counts split apart, and that split is the reason this tool exists. The sample string it opens with is a case in point: 149 graphemes, 149 code points, 150 code units, and 152 bytes, all describing the same one sentence with a single emoji at the end of it.

The count you care about depends entirely on who is counting. A grapheme is what Unicode Standard Annex 29 calls an extended grapheme cluster, and it is the closest formal definition of what an ordinary person means by one character. A code point is one entry in the Unicode catalogue.

A UTF-16 code unit is a storage slot in the encoding that JavaScript, Java, and Windows use internally, which is why an astral character costs two of them and why a plain call to string length in a browser reports an emoji as 2.

A UTF-8 byte is what actually travels down the wire or sits in a database column. A family emoji reads as 1, 7, 11, and 25 on those four scales respectively, and none of those numbers is wrong.

Nothing you paste leaves your browser. The counting runs in the page itself: no upload, no server round trip, nothing stored between visits, and the text is never written into the page URL, so there is no shareable link quietly carrying your draft around.

That matters when the string is an unpublished headline, a client message, a password hint, or a paragraph from a document you are not free to share. The platform selector is only a display choice, not a filter, so switching between the SEO title limit and the X limit changes the progress reading and the accompanying note while the four underlying counts stay exactly where they were.

Why does a post that should fit get rejected?

That spread is why a post that should have fitted gets rejected. Take twenty-six family emoji: you counted twenty-six characters, and you are right by any human measure, but the same string is 286 UTF-16 code units.

A form that validates on storage length sees 286 against a limit of 280 and refuses it, with an error message that will tell you it is 6 over and never tell you why, because nothing in the visible text looks any longer than the twenty-six symbols you can point at.

The formula, worked line by line

There is no arithmetic here, which is the point: the difficulty is not calculating a count but deciding which count to report. This tool reports four, side by side, because the four are genuinely different questions and every one of them is the right answer somewhere.

Each comes from a different mechanism. The grapheme count uses the segmenter built into the browser, which applies the extended grapheme cluster rules of Unicode Standard Annex 29. The code-point count walks the string one Unicode scalar at a time. The code-unit count is the raw length property, the number most code reports without thinking. The byte count encodes the string as UTF-8 and measures the result.

graphemes      = Intl.Segmenter(locale, { granularity: "grapheme" })
code points    = [...text].length
code units     = text.length          (UTF-16, what JavaScript reports)
bytes (UTF-8)  = TextEncoder().encode(text).length
no spaces      = code points, with all whitespace removed

The platform limits offered on this page are current as of July 2026. Platforms change these, so treat them as a starting point rather than a permanent fact and verify against the platform documentation before anything expensive depends on one; the notes attached to each limit below describe what actually happens at the edge, which is often softer than a hard refusal.

• SEO title tag: 60 characters, where Google truncates on pixel width near that mark rather than on a hard count.
• Meta description: 155 characters, which Google rewrites more often than not but which remains the safe span.
• X post: 280 characters, where a link counts as 23 characters no matter how long it really is.
• SMS on the GSM 7-bit alphabet: 160 characters, where a single character outside that alphabet drops the whole message to the Unicode ceiling.
• SMS in Unicode: 70 characters, which any emoji or non-Latin character puts you into.
• Instagram caption: 2,200 characters, of which only about the first 125 show before the more link.
• LinkedIn post: 3,000 characters, truncated at roughly 210 in the feed.
• YouTube title: 100 characters, cut around 70 in search results.
• Facebook post: 63,206 characters, where engagement falls off long before the ceiling.

Two practical rules follow. First, when a form rejects text that you counted as fitting, look at the code-unit column: the culprit is almost always an emoji, a flag, or a skin-tone modifier, and removing one can buy back several units at once.

Second, when the constraint is storage rather than display, count bytes. A database column declared to hold 20 characters in a byte-measured encoding holds five Japanese kana comfortably at 15 bytes but chokes on a single family emoji at 25. The same applies to filenames, HTTP headers, and anything with a documented byte budget.

Choosing the wrong count is not a small error at the margin; on emoji-heavy text the counts differ by more than ten to one.

Questions people ask