Remove zero-width and invisible characters

The cleaner recognizes a deliberately narrow set of code points so its report is explainable. These include zero-width space, word joiner, embedded byte-order marks, Unicode tag characters, supported bidirectional embedding or isolate controls, and non-standard spacing characters that can be normalized to an ordinary space. Counts are shown before any change is applied.

Zero-width joiner and zero-width non-joiner are different: they can shape words in some writing systems and combine emoji into a single visible glyph. The safe preset therefore keeps them. The interface should not call all invisible code points malicious, because direction controls, variation selectors, combining marks, and joiners can have legitimate linguistic or typographic roles.

A hidden Unicode character is part of a text string. It can affect searching, comparison, cursor movement, line wrapping, direction, or rendering even though it has little or no visible width. Cleaning it changes the sequence of code points, not metadata attached to a separate file.

Image metadata lives in EXIF, XMP, IPTC, C2PA, or format-specific chunks, while an invisible image watermark is encoded into pixels. This text page handles neither. Pasting an image filename, an image URL, or text extracted by OCR does not inspect the underlying image container.

Removing bidi controls can change the displayed order of mixed right-to-left and left-to-right text. Normalizing spaces can alter line wrapping or byte-for-byte signatures. Joiners and selectors may be invisible but semantically important. Review multilingual, source-code, domain-name, legal, and cryptographically signed material with the appropriate specialist tools before replacing the original.

This cleaner can expose and remove supported characters, but it is not a malware scanner, homoglyph detector, plagiarism checker, AI detector, or authorship verifier. It does not promise that cleaned text is safe, original, human-written, or accepted by a platform.