Invisible Character Remover

Using the tool

• Reveal shows the text with each hidden character as a badge: red will change, grey is kept. Select a badge to switch that one character.
• Cleaned shows the result; Copy copies it from either view.
• Characters found lists each character with its code point, what it does and how often it appears. Where some occurrences are in use, such as joiners inside an emoji, they get a row of their own, kept by default. A row's checkbox switches all of its occurrences; Select all selects every row, and Reset returns to the defaults.
• Editing the text resets every choice to the defaults, since an edit moves the characters the choices were made on.

Most characters are removed. No-break and other unusual spaces become a plain space, so words stay apart. Vertical tab, form feed, next line and the Unicode line and paragraph separators become a line break of the kind the text already uses.

Each example below starts with the text as pasted, hidden characters included: copy it into the tool to see the same result.

What it finds and what it keeps

The tool reports every character Unicode classes as default-ignorable (meant to be invisible when not supported), every space other than the plain space, tab and line breaks, every control character other than tab and line breaks, and broken halves of encoded pairs. These come from the browser's own Unicode data, so a newer browser recognises characters added in newer Unicode versions.

Some of these characters do a job, and are kept unless you select them:

• A zero width joiner between two emoji, as in a family or a profession emoji, or between letters of a script that uses joiners, such as Hindi, Malayalam or Arabic.
• A zero width non-joiner between letters of such a script, as in Persian, where it keeps two letters from joining.
• A variation selector that forms a sequence Unicode lists, such as the emoji style of ❤, and a single ideographic selector after a Chinese, Japanese or Korean character.
• The tag characters of the flags of England, Scotland and Wales, the only flags written with them.
• Direction marks and controls in text that has Arabic, Hebrew or another right-to-left script. A direction override is always selected, since it can make text display in a different order than it is stored, along with the character that ends it.
• A zero width space between Thai, Lao, Khmer or Myanmar letters, which those scripts use to mark where words end; an ideographic space beside Chinese, Japanese or Korean text; and the Mongolian and Hangul format characters inside their own scripts.

Examples

Hidden characters by name

Standards

Which characters count follows the Unicode Character Database: the Default_Ignorable_Code_Point and White_Space properties and the control category. Variation sequences come from Unicode's standardized and emoji variation sequence lists, flag tag sequences from the Unicode emoji standard, and the pairing of direction controls from the Unicode bidirectional algorithm. Direction overrides that disguise source code are the "Trojan Source" vulnerability, CVE-2021-42574.

AI text watermarks, as Google DeepMind and OpenAI describe theirs, are patterns in the words a model picks rather than characters, so they are outside what this tool can find.

Sources

• DerivedCoreProperties.txt (Default_Ignorable_Code_Point) — Unicode Consortium
• PropList.txt (White_Space) — Unicode Consortium
• StandardizedVariants.txt — Unicode Consortium
• emoji-variation-sequences.txt — Unicode Consortium
• UAX #9: Unicode Bidirectional Algorithm — Unicode Consortium
• UTS #51: Unicode Emoji — Unicode Consortium
• CVE-2021-42574 — NIST National Vulnerability Database
• SynthID — Google DeepMind
• Understanding the source of what we see and hear online — OpenAI

Reviewed 8 October 2026

Frequently asked questions

See also

• Remove Extra Spaces — to collapse repeated spaces and drop blank lines next.
• Find and Replace — to replace a specific character everywhere it appears.
• Word & Character Counter — to count what remains.
• Text to Binary Converter — to see the bytes of each character.