Japanese Character Count on Twitter: Why CJK Costs Double on X

The Japanese character count on Twitter is not what a character counter tells you. X weights each CJK character as two rather than one, so the practical limit for a Japanese post is 140 characters, not 280 — and a Japanese caption that is genuinely shorter than the English can still be the one that overflows.

X does not count characters. It counts weight, and it publishes the rule in its developer documentation on counting characters.

Most characters weigh 1. Characters in a set of ranges covering CJK ideographs, kana, Hangul and full-width forms weigh 2. Every URL weighs a flat 23, whatever its real length. The 280 you are budgeting against is a weight budget wearing a character budget's name.

It is not a penalty. It is an attempt to make the limit mean the same thing in every script.

A Japanese character is roughly twice the width of a Latin one on screen, and it also carries far more meaning — a sentence that takes sixty characters in English often takes twenty-five in Japanese. Counting each one as two makes the limit approximate a fixed amount of screen space and a fixed amount of content rather than a fixed number of glyphs, which is closer to what a timeline actually constrains.

The consequence is that Japanese posters are not disadvantaged. 140 Japanese characters carry more than 280 English ones. The rule only surprises people who assumed a character was a character.

A caption is translated from English into Japanese. It comes back 40% shorter in characters. It is over the limit.

Nothing is wrong with the translation. It is shorter by every measure a person would use, and it costs more of the budget, because 150 characters at weight 2 is 300. This is the single most common way a multilingual account discovers the weighting rule.

It shows the four numbers side by side — bytes, codepoints, graphemes and X's weighted count — and breaks down exactly which characters produced the gap.

Chinese behaves the same way and shortens more. A Chinese translation is frequently half the character count of the English, which at weight 2 lands at parity.

Korean is weighted at 2 for Hangul syllable blocks, and Korean word spacing works more like Latin script than the other two, so the count sits between them.

Thai, Hindi, Arabic and Hebrew are not weighted. They count 1 per character like Latin. Arabic and Hebrew bring a different problem — a Latin handle or URL inside right-to-left text renders in an order that looks wrong on first sight — but the budget is the ordinary one.

Three practical things.

Budget the English at half. If a post has to work in Japanese, the English version has 140 weighted characters to play with, not 280. Working that out afterwards means cutting the translation.

Do not shorten links to save room. Every URL costs 23 whatever its length, so a shortened link is charged the same as the long one — and a link already under 23 characters is charged more than it costs. Social media short links covers when shortening is worth doing, and it is never for X's character budget.

Check with a counter that applies the rule. An editor's word count, a text field's maxlength and a generic character counter will all tell you the post fits. Only a counter applying X's own weighting will tell you whether it does.

The wider version of this problem — a caption that fits in the source language and not in the target — is in text expansion in translation, and the overall picture is in translating social media posts.

Paste the post into the weighted character counter to see the weighted total beside the one your editor is showing you.