X's developer page Counting Characters says a Post can hold 280 characters and that not every character counts the same. The counting is done by twitter-text, the open-source library the page points to, which gives each code point a weight of 1 or 2 against the limit. We ran release 3.1.0 on Article 1 of the Universal Declaration of Human Rights from Unicode's UDHR in Unicode collection. English counted 170. Khmer counted 372. The article fits in one Post in English, Thai, Burmese, Tamil, Korean and Chinese, and runs over in Khmer and in Javanese script. Across the whole declaration a visible Khmer character counted about 3.5 times an English one.
Four ranges of code points form the foundation of the library's count
The weights live in one file, config/v3.json, unchanged since October 2018. A code point counts 1 inside a listed range and 2 everywhere else. Four ranges are listed: U+0000 to U+10FF, then three short runs of punctuation, U+2000 to U+200D, U+2010 to U+201F and U+2032 to U+2037. The same file fixes every link at 23.
The first range decides scripts. It covers Latin, Cyrillic, Arabic, Tamil, Thai, Lao and Burmese, and stops just short of Hangul. The punctuation runs explain oddities: curly quotation marks count 1 while the ellipsis and the euro sign count 2, and an ASCII comma counts 1 where the full-width , counts 2.
X's page gives a shorter table, with "Other Unicode" at 2; the configuration folder it links to sets the exact ranges, so every per-script figure here is the library's.
Khmer is the only mainland script laid wholly in the upper course
A visible character here is Unicode's default extended grapheme cluster, counted with Intl.Segmenter in Node 26.13.0 (ICU 79.1, Unicode 18.0), and a whole text is every paragraph in the file, prefatory note included. The texts are counted, never quoted, as their licence is unconfirmed.
| Text | Article 1, visible characters | Article 1, library count | Whole text, count per visible character |
|---|---|---|---|
| English | 170 | 170 | 1.00 |
| Malay, Latin script | 200 | 200 | 1.00 |
| Malay, Jawi | 161 | 161 | 1.00 |
| Vietnamese | 183 | 206 | 1.13 |
| Thai | 115 | 144 | 1.25 |
| Lao | 110 | 148 | 1.32 |
| Tamil | 157 | 238 | 1.57 |
| Burmese | 159 | 262 | 1.64 |
| Korean | 87 | 153 | 1.73 |
| Chinese, Simplified or Traditional | 43 | 86 | 1.97 |
| Khmer | 101 | 372 | 3.54 |
| Javanese script | 71 | 298 | 4.01 |
For mainland Southeast Asia, the results split three to one. Thai, Lao and Burmese count 1 per code point, so their ratios above 1 only measure how many code points make each visible character. Khmer starts at U+1780 and counts 2 throughout, coeng sign ្ included. Its clusters average about 1.8 code points, and at 2 apiece that comes to roughly the 3.5. In this segmenter a coeng and the consonant written beneath it stay in one cluster. Our guide on what a character limit does to Khmer covers clusters.
The first 280 units of the whole declaration hold 223 visible Thai characters and 85 Khmer ones. Jawi, the Arabic-script spelling of Malay, counted exactly 1.00 over 9,414 characters. Javanese script counted higher than Khmer.
Vietnamese letters from a later Unicode block are bricked in beside Khmer
Vietnamese uses Latin letters, yet counted 1.13 against Malay's 1.00, because of one Unicode block. Letters such as ế and ữ are encoded precomposed in Latin Extended Additional, U+1E00 to U+1EFF, and count 2. Letters encoded earlier, among them é, ơ, ư and đ, count 1.
In the whole Vietnamese text, 1,364 of 10,729 visible characters (12.7%) come from the later block. About a quarter of those, 347, carry a single mark, as ả does, so stacked accents are only part of it.
X's page says length is calculated after Unicode Normalization Form C. The collection's Vietnamese file is stored decomposed, 12,641 code points composing to 10,729 characters, and the total is the same either way. The Vietnamese section of why your name does not fit in the box explains the two forms.
Emoji newer than the library's 2018 list are laid code point by code point
X's page states that "All emojis count as 2 characters, regardless of complexity". The library finds emoji with twemoji-parser 11.0.2, published in October 2018, and anything on that list counts 2, as the family 👨👩👧👦 and the flag 🇸🇬 do.
A sequence added later is counted code point by code point. In our run the phoenix 🐦🔥 counted 5 and the handshake with two skin tones 🫱🏻🫲🏿 counted 9. A newer emoji that is one code point, such as 🫠, still counted 2, so a quick spot check passes. That is the 2020 library beside the page's rule; X's own composer we did not test. Links agree at 23 in both.
JavaScript string length sees half the bricks in Chinese and extra ones in emoji
A counter built on JavaScript's .length counts UTF-16 code units. On Chinese Article 1 it gives 43 where the library gives 86, and on Khmer it lands near half the library's count for the same reason. Emoji run the other way: the family emoji is 11 units against the library's 2, and a link counts at its real length. Counting graphemes instead fixes neither direction, since Chinese Article 1 is 43 of those too.
This generator builds usernames to each platform's published rules and leaves length alone
Our username and bio generator makes usernames and a display name from a keyword, and adds a bio line. It checks each username against the published rules of Instagram, TikTok, X, YouTube, Snapchat and Discord, and its source code names where and when each rule was read. For Instagram it checks characters only, because the page it read states no length limit. It checks no bio length. Each bio is one of four fixed lines per style and language, all short enough for Instagram's 150.
Until 9 November 2026 its keyword cleaner deleted letters that do not decompose to plain Latin, so the surname Đặng came out as "ang". Its Chinese bios used ASCII commas, and its Discord check passed hyphenated names that the Discord Support article "New Usernames & Display Names" refuses. The revision of that date maps đ to d, giving dang, uses full-width Chinese punctuation, and applies the article's list: lowercase a to z, digits, underscore and period.
Only Instagram's bio limit appears in a blueprint we could read
Instagram's help page Add a bio to your Instagram profile says "You can write a bio of up to 150 characters", the one bio limit stated on a page we read. It does not say how a character is counted. Two Threads help pages give no length. X's help pages sat behind a browser challenge our requests could not pass, and TikTok's returned no text without a browser.
Everything measured above concerns Posts. X files its weighting under how it counts characters in Posts, and nothing we could read says a profile bio is counted the same way. A Khmer writer can now say how far Article 1 runs past one Post under X's library, and still has no published figure for how much Khmer an X bio will take.