Control character filter

A control character does not represent a printable letter or symbol. It encodes an instruction: join two characters, force a run of text to flow right to left, mark a boundary a line wrap should respect. None of that produces a rendered glyph, while all of it remains present as real bytes in the text.

This tool scans pasted text for that category and removes it without repositioning or altering any visible character. Detection and removal run entirely client-side, so there is no upload, no server round trip and no run limit to speak of.

Nothing to inspect beforehand, since these characters are non-printing by definition.

Step 1

Paste the text into the input field.

Step 2

Run the filter. It executes instantly, locally.

Step 3

Read the result; every visible character is unaffected.

Unicode assigns every codepoint to a general category. Letters, digits and punctuation fall into categories a font renders as a visible shape. A distinct category, labelled Format, is reserved for characters with no rendered shape at all: their function is to modify the treatment of surrounding characters, not to be displayed themselves.

That Format category is what this tool targets: zero-width joiners and non-joiners, the word joiner, the soft hyphen, and the directional formatting characters governing left-to-right versus right-to-left text flow.

U+202D and U+202E do not merely fail to render. They actively reverse the display order of every character that follows, until a matching pop-directional character or the end of the text cancels the effect.

This has a documented misuse: a right-to-left override embedded in a file name can make a string like 'invoice_gpj.exe' display as 'invoice_exe.jpg', reversing the apparent extension while leaving the underlying file unchanged. This tool removes these override characters along with the rest of the Format category.

A control character has no rendered shape, so nothing on screen prompts a person to notice or remove it during editing. A copy operation transfers the underlying string exactly as it is, invisible characters included, and most editors give no visual indication that anything beyond the rendered letters travelled with it.

That is the mechanism behind why this class of defect resists visual inspection: a message, file name or field all present as entirely normal, until something downstream reads the raw byte sequence rather than the rendered output.

Logging systems and terminals can render control characters as stray output, or in some cases interpret one as an escape sequence, altering how subsequent output is displayed. A file name or URL carrying a directional override can present as something other than what it is, a documented technique for disguising a malicious attachment or link.

In source code, a control character embedded in a string literal or identifier produces an error with no visible trace of its cause in an editor. In structured data, one embedded in a field can break strict parsing outright.

Security-conscious reviewers check suspicious file names, URLs or pasted messages here for directional-override tricks before opening or forwarding them. Engineers filter externally sourced text before it reaches a logging pipeline, a terminal or a parser that could misinterpret a stray control character as an instruction rather than data.

Teams operating systems that store or render user-submitted text use it to strip formatting characters that could otherwise alter how that text later displays to another viewer.