Paste a string and the tool takes it apart: every visible character, the code points behind it, and a list of anything invisible hiding between the letters. It is the fastest way to find out why two strings that look identical will not match.
That last case is the practical one. A name that will not validate, a coupon code that "is definitely right", an email address that bounces โ often the text contains a zero-width space, a no-break space or a soft hyphen that arrived with a copy and paste and never shows on screen.
Three different counts
The summary reports three numbers, and they are not interchangeable:
| Unit | What it is | a๐๐ฝb |
|---|---|---|
| Graphemes | What a reader sees as one character | 3 |
| Code points | The underlying Unicode values | 4 |
| UTF-16 units | How JavaScript stores the string | 6 |
That example is the whole lesson. Three visible characters, four code points, six storage units โ because the thumbs up is one code point stored in two units, and the skin tone modifier is a second code point that joins it to make a single visible character.
Character limits on social platforms and forms count differently, and that is how a bio that looks 150 characters long gets rejected. The ground truth for a limit is usually the UTF-16 or code point count, not the grapheme count. This tool reports all three so you can compare.
Finding invisible characters
Each code point is checked against a table of about 36 known invisible and formatting characters. Anything found is marked with a tone, labelled by name in the grid, and counted in the status line. The names that matter most:
- U+200B zero-width space โ the usual culprit after a copy from a web page or a PDF.
- U+00A0 no-break space โ looks like a space, is not one, and survives most trimming.
- U+FEFF byte order mark โ the invisible character at the start of a file exported as UTF-8.
- U+00AD soft hyphen โ invisible until a line break happens at that point.
- U+200D zero-width joiner โ legitimate inside emoji, suspicious in a name or a code.
The grid gives the code points for every grapheme, so a combined character shows as U+1F468 U+200D U+1F469 U+200D U+1F467 rather than as one opaque symbol. That is how you tell a real family emoji from three separate people and two joiners.
The hidden count has false positives
This is the part to know before you trust the amber warning. The check that flags control characters compares a hexadecimal string against the decimal number 32, so any code whose hex value reads as a number below 32 is treated as hidden. That catches the real control characters, and it also catches eleven perfectly ordinary ones:
Hello, world!
Reported:
2 hidden characters
Actually hidden:
none โ the two are the space and the exclamation markThe full set of false positives is the space, !, ", #, $, %, &, ', (, ), 0 and 1. In practice that means almost any English sentence reports hidden characters, because it contains a space or a zero. The loaded sample reports 7 hidden characters, and four of them are the digits and the space in "price: 100".
Read the codes rather than the count. Use the hidden filter, then look at each entry: a real invisible character has a code in the U+2000 to U+206F range, or U+00A0, U+00AD, U+FEFF or U+034F. If the code is U+0020 to U+0031, it is an ordinary character the check misread. Punctuation from U+002A onwards is not affected, which is why a comma is treated correctly and a space is not.
What the emoji count misses
The emoji filter looks for characters in the Extended Pictographic property. That covers faces, hands, animals and most objects, and it misses two common groups:
- Flags. A flag is a pair of regional indicator letters, which are not pictographic, so a British or Indian flag is counted as a combined character but not as an emoji.
- Keycaps. A keycap like 1๏ธโฃ is a digit plus a variation selector plus an enclosing keycap. None of the three is pictographic, so it is counted as hidden instead.
Skin tone modifiers and zero-width joiners also count as hidden rather than as emoji, because they are combining marks. On the emoji sample, Hi ๐จโ๐ฉโ๐ง ๐ฌ๐ง ๐๐ฝ reports 8 graphemes, 14 code points, 21 UTF-16 units and 2 emoji โ the family and the thumbs up, with the flag left out.
Views and grouping
Three views filter the grid: all characters, hidden only, or emoji only. On a long text the grid stops at 240 graphemes and says how many it is showing, so a pasted document is inspected in part โ the counts above the grid always describe the whole string.
Group repeats folds consecutive identical characters into one row. The note on a grouped row lists every repeat count rather than the final one, so four "a" characters in a row show as "ร2 ยท ร3 ยท ร4". It is readable, if clumsy, and it only groups runs of the same character โ "abab" stays four rows.
Three good reasons to paste something here
- A form rejects a value that looks correct. Find the zero-width or no-break character, delete it, resubmit.
- A string comparison fails in code. Check whether one side carries a combining mark or a different apostrophe.
- A character count does not match the platform's. Compare graphemes, code points and UTF-16 units to see which one the limit is using.
The copy button produces a plain-text list of the first 80 graphemes with their codes, which is the format to paste into a bug report. Nothing is uploaded and a reload clears the box.