Invisible characters

Copy a blank character for a name or a message, or paste text to find the ones hiding in it.

Paste text to look inside 0 / 20,000

Try

See through them in your app, with one call

Every invisible character, ready to copy

Each with its code point, the way to write it in HTML and a button that copies it.

Character	Code point	HTML	What it does
Zero-width characters No width at all: the ones people mean by an invisible character.
Zero width space ZWSP	U+200B		No width at all. The usual way to split a word a filter is looking for.
Zero width non-joiner ZWNJ	U+200C	‌	Keeps two letters from joining. Persian and several Indian scripts need it.
Zero width joiner ZWJ	U+200D	‍	Joins emoji into one, as in 👩‍💻, and letters in some scripts.
Word joiner WJ	U+2060	⁠	A zero width space that also forbids a line break.
Zero width no-break space BOM	U+FEFF		The byte order mark. At the start of a file it is metadata; anywhere else it hides.
Blank letters Drawn as an empty letter. Games and chat apps that refuse a blank name often take these.
Hangul filler	U+3164	ㅤ	A blank Korean letter. The usual blank name in games and chat.
Hangul choseong filler	U+115F	ᅟ	A blank Korean initial consonant.
Hangul jungseong filler	U+1160	ᅠ	A blank Korean vowel.
Halfwidth Hangul filler	U+FFA0	ﾠ	The half-width blank Korean letter.
Braille pattern blank	U+2800	⠀	A braille cell with no dots: a blank that counts as a letter.
Format characters Formatting with nothing to draw, most of them deprecated.
Soft hyphen SHY	U+00AD		Shows a hyphen only where a line breaks; invisible mid-line.
Combining grapheme joiner CGJ	U+034F	͏	A combining mark that draws nothing.
Mongolian vowel separator MVS	U+180E	᠎	A space until Unicode 6.3; now it draws nothing.
Khmer vowel inherent aq	U+17B4	឴	A Khmer vowel Unicode deprecated; it draws nothing.
Khmer vowel inherent aa	U+17B5	឵	A Khmer vowel Unicode deprecated; it draws nothing.
Function application	U+2061	⁡	An invisible maths operator.
Invisible times	U+2062	⁢	An invisible maths operator.
Invisible separator	U+2063	⁣	An invisible maths comma.
Invisible plus	U+2064	⁤	An invisible maths operator.
Inhibit symmetric swapping	U+206A	⁪	A deprecated shaping control; it draws nothing.
Activate symmetric swapping	U+206B	⁫	A deprecated shaping control; it draws nothing.
Inhibit Arabic form shaping	U+206C	⁬	A deprecated shaping control; it draws nothing.
Activate Arabic form shaping	U+206D	⁭	A deprecated shaping control; it draws nothing.
National digit shapes	U+206E	⁮	A deprecated shaping control; it draws nothing.
Nominal digit shapes	U+206F	⁯	A deprecated shaping control; it draws nothing.
Direction controls They set which way text runs.
Left-to-right mark LRM	U+200E	‎	Sets the direction of the text around it. Invisible.
Right-to-left mark RLM	U+200F	‏	Sets the direction of the text around it. Invisible.
Arabic letter mark ALM	U+061C	؜	A right-to-left mark for Arabic text.
Left-to-right embedding LRE	U+202A	‪	Starts a run of left-to-right text.
Right-to-left embedding RLE	U+202B	‫	Starts a run of right-to-left text.
Pop directional formatting PDF	U+202C	‬	Ends an embedding or an override.
Left-to-right override LRO	U+202D	‭	Forces the text after it left to right.
Right-to-left override RLO	U+202E	‮	Shows the text after it backwards: "gpj.exe" reads as "exe.jpg".
Left-to-right isolate LRI	U+2066	⁦	Starts an isolated left-to-right run.
Right-to-left isolate RLI	U+2067	⁧	Starts an isolated right-to-left run.
First strong isolate FSI	U+2068	⁨	Starts a run in the direction of its first letter.
Pop directional isolate PDI	U+2069	⁩	Ends an isolate.
Odd spaces They look like a space and are not one, so a search or a split misses them.
No-break space NBSP	U+00A0		Looks like a space, but a word on either side never splits.
En quad	U+2000		A space half an em wide.
Em quad	U+2001		A space an em wide.
En space	U+2002		A space half an em wide.
Em space	U+2003		A space an em wide.
Three-per-em space	U+2004		A third of an em.
Four-per-em space	U+2005		A quarter of an em.
Six-per-em space	U+2006		A sixth of an em.
Figure space	U+2007		The width of a digit.
Punctuation space	U+2008		The width of a full stop.
Thin space	U+2009		A narrow space.
Hair space	U+200A		The narrowest space.
Narrow no-break space NNBSP	U+202F		A thin space that never breaks.
Medium mathematical space MMSP	U+205F		A space for formulas.
Ideographic space	U+3000		A space as wide as a Chinese character.
Line separator	U+2028		A line break some apps honor and some ignore.
Paragraph separator	U+2029		A paragraph break some apps honor and some ignore.

How invisible characters slip past a filter

Unprofane takes each one out, or reads it as its Latin letter, before it reads the words.

A zero width space insideshZWSPit shitreviewzero_width99.7% caught
Full-width lettersｆｕｃｋ off fuck offblockfullwidth100.0% caught
A Cyrillic letterѕhithead shitheadblockhomoglyph99.6% caught
A mark under every letterf̲u̲c̲k̲ off fuck offblockcombining_marks100.0% caught

Caught, by trick: the Scunthorpe Test, rules alone, 2026-10-20

Your users hide words this way

Every check reads through them before it reads the words, and tells you which trick it found.

Get your API key

100 checks a day, no card

curl -X POST https://api.unprofane.com/v1/check \
  -H "Authorization: Bearer up_test_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text":"nice game, f\u200buck you","context":"chat"}'

const res = await fetch('https://api.unprofane.com/v1/check', {
  method: 'POST',
  headers: { Authorization: 'Bearer up_test_YOUR_KEY', 'Content-Type': 'application/json' },
  body: JSON.stringify({
    "text": "nice game, f\u200buck you",
    "context": "chat"
  }),
});
const result = await res.json();

import httpx

r = httpx.request(
    "POST",
    "https://api.unprofane.com/v1/check",
    headers={"Authorization": "Bearer up_test_YOUR_KEY"},
    json={
        "text": "nice game, f\u200buck you",
        "context": "chat",
    },
    timeout=5.0,
)
result = r.json()

req, _ := http.NewRequest("POST", "https://api.unprofane.com/v1/check",
	strings.NewReader(`{"text":"nice game, f\u200buck you","context":"chat"}`))
req.Header.Set("Authorization", "Bearer up_test_YOUR_KEY")
req.Header.Set("Content-Type", "application/json")
res, err := http.DefaultClient.Do(req)

$res = file_get_contents('https://api.unprofane.com/v1/check', false, stream_context_create(['http' => [
  'method' => 'POST',
  'header' => "Authorization: Bearer up_test_YOUR_KEY\r\nContent-Type: application/json\r\n",
  'content' => '{"text":"nice game, f\u200buck you","context":"chat"}',
]]));
$result = json_decode($res, true);

require 'net/http'; require 'json'
uri = URI('https://api.unprofane.com/v1/check')
req = Net::HTTP::Post.new(uri, 'Authorization' => 'Bearer up_test_YOUR_KEY', 'Content-Type' => 'application/json')
req.body = '{"text":"nice game, f\u200buck you","context":"chat"}'
res = Net::HTTP.start(uri.host, uri.port, use_ssl: true) { |http| http.request(req) }
result = JSON.parse(res.body)

using var http = new HttpClient();
http.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", "up_test_YOUR_KEY");
var req = new HttpRequestMessage(HttpMethod.Post, "https://api.unprofane.com/v1/check");
req.Content = new StringContent("{\"text\":\"nice game, f\u200buck you\",\"context\":\"chat\"}", Encoding.UTF8, "application/json");
var res = await http.SendAsync(req);
var result = await res.Content.ReadFromJsonAsync<JsonElement>();

var http := HTTPRequest.new()
add_child(http)
http.request_completed.connect(func(_r, _code, _h, body):
	var result = JSON.parse_string(body.get_string_from_utf8())
	print(result))
http.request("https://api.unprofane.com/v1/check",
	PackedStringArray(["Authorization: Bearer up_test_YOUR_KEY", "Content-Type: application/json"]),
	HTTPClient.METHOD_POST, "{\"text\":\"nice game, f\u200buck you\",\"context\":\"chat\"}")

Response200

{
  "allowed": false,
  "action": "block",
  "evasion": [
    "zero_width"
  ],
  "matches": [
    {
      "text": "f\u200buck you",
      "normalized": "fuck you",
      "evasion": [
        "zero_width"
      ]
    }
  ],
  "censored": "nice game, *********"
}

Invisible characters, answered

What is an invisible character?

An invisible character takes up a place in text but draws nothing, or draws a blank: a zero width space, a blank Korean letter, a braille cell with no dots. Apps count it as a character, so it can fill a name that may not be empty, or sit inside a word with no gap to see. The table has every one, each with a copy button.

How do I copy a blank character for a name?

Press Copy beside the Hangul filler (U+3164) at the top of this page, then paste it where the name goes. It is a blank Korean letter, the usual blank name in games and chat. Where an app trims it or draws a box, try the Braille pattern blank (U+2800).

How do I remove invisible characters from text?

Paste the text into the detector and press Copy clean text. The clean copy has the invisible characters taken out, odd spaces made plain and look-alike letters made Latin, and you can turn each of those off. Emoji joiners and the joiners a script needs stay unless you say otherwise, and nothing leaves your browser.

Can invisible characters get a word past a profanity filter?

Yes, past a filter that matches letters as typed: a zero width space inside a word leaves two harmless pieces. Unprofane takes them out before it reads the words and names the trick; on the Scunthorpe Test's public split the rules alone caught 99.7% of words split by zero-width characters (run of 2026-10-20).

Can invisible characters hide instructions for an AI model?

Yes. Unicode tags spell ASCII without drawing it, so a page, an email or a pasted prompt can carry a line a person never sees and a model reads in full. The detector reads the line out; strip tags from anything you pass to a model.

What is a right-to-left override?

An invisible control (U+202E) that shows the text after it reversed. A file named invoice_2026[RLO]fdp.exe reads as a PDF and runs as a program, and in source code the same controls can make a line read differently from how it compiles. The detector marks each one where it stands.

Which invisible characters are not in the table?

The variation selectors and the Unicode tags, named by range rather than listed. The 256 variation selectors (U+FE00 to U+FE0F and U+E0100 to U+E01EF) pick a form of the character before them; the tags (U+E0000 to U+E007F) are an invisible copy of every ASCII character. The detector reads both, and reads out any text spelled in them.

Every invisible character, ready to copy

How invisible characters slip past a filter

Your users hide words this way

Invisible characters, answered

More free tools