Tweet Character Counter

A standard post is limited to 280 weighted characters. Weighted is the important word: X does not count code points, it assigns each one a weight and compares the total against 280. Most Latin, Cyrillic and Greek text, plus general punctuation, weighs 1 per character. Everything outside those ranges, which includes emoji and CJK writing, weighs 2. Accounts with a paid subscription can post far longer, up to 25,000 characters.

The published weighting configuration gives a default weight of 200 and lowers it to 100 only for four code point ranges. Those ranges cover the Latin, Cyrillic and Greek blocks plus some punctuation. Emoji sit outside them, so they take the default weight, which after the scale of 100 works out to 2 per character. The same rule is why Japanese, Chinese and Korean text hits the limit at roughly half as many visible glyphs.

Yes, both count as ordinary text at their real weighted length, unlike links. A mention of a fifteen-character handle costs sixteen characters including the at sign. The one historical exception people remember, replies not counting the handles they are addressed to, applied to the leading mentions in a reply thread and is not something a counter can infer from pasted text alone.

Every link is replaced by a fixed transformed length of 23, whatever the real URL is. A six-character short link and a two-hundred-character tracking URL both cost the same 23. That is why the raw character count in a text editor and the count in the composer disagree the moment a link appears, and why shortening a URL by hand buys you nothing at all.

It implements the same published weighting rules: four light ranges at weight 1, everything else at 2, links fixed at 23. It is an independent implementation rather than the library X ships, so the composer has the final say on unusual grapheme clusters such as multi-code-point emoji sequences with skin tone or gender modifiers. For ordinary text, links, mentions and hashtags the two agree.

Apply the same rules in your own code: iterate code points, add 1 for anything in the four light ranges and 2 otherwise, then replace each URL match with a flat 23. If you are counting posts you have fetched rather than posts you are writing, the tweet endpoints return the post body and its entities, so you can measure real posts at scale instead of guessing from a sample.