About

Text corruption exploits the Unicode combining character mechanism defined in the Unicode Standard (Chapter 3, Conformance D52). Combining diacritical marks in ranges U+0300 - U+036F and U+1DC0 - U+1DFF attach to any preceding base character without advancing the cursor position. Stacking n combining marks on a single base glyph forces the rendering engine to draw overlapping marks above and below the baseline, producing the visual distortion known as "Zalgo" text. Misconfigured text sanitizers that strip only ASCII will pass these through, which is why corrupted text appears intact on most platforms. This generator also applies homoglyph substitution - replacing Latin a (U+0061) with Cyrillic а (U+0430) - a technique documented in Unicode Technical Report #36 (Security Considerations). Intensity is controlled probabilistically: each base character has a corruption probability of I100, where I is the slider value.

Limitation: rendering varies across platforms. Some mobile keyboards collapse excessive combining marks. Discord and Twitter may strip certain ranges. Test your output in the target application before committing to a style. This tool operates entirely client-side - no text is transmitted to any server.

Formulas

The corruption probability for each character at position i is governed by a uniform random variable compared against the normalized intensity parameter:

P_corrupt = I100

where I ∈ [0, 100] is the intensity slider value. A character is corrupted when Math.random() < P_corrupt.

For Zalgo-style corruption, the number of combining marks appended to a base character c is:

n_marks = ceil(I10) + floor(Math.random() × ceil(I5))

where n_marks determines the vertical extent of the distortion. At I = 100, a single base character may carry up to 30 combining marks.

Homoglyph substitution uses a bijective mapping H: L → U, where L is a subset of Basic Latin (U+0041 - U+007A) and U contains visually confusable codepoints from Cyrillic, Greek, or Mathematical Alphanumeric blocks. The mapping is deterministic per character but application is stochastic, controlled by P_corrupt.

Reference Data

Corruption Style	Unicode Range / Technique	Visual Effect	Platform Support
Zalgo (Above)	U+0300 - U+036F (Combining Diacriticals)	Marks stack above glyphs	Universal
Zalgo (Below)	U+0316 - U+0333 (Below Diacriticals)	Marks drip below baseline	Universal
Zalgo (Overlay)	U+0334 - U+0338 (Combining Overlays)	Strikes through glyph center	Universal
Homoglyph	Cyrillic аеорс → Latin a e o p c	Visually identical, different codepoint	Universal
Fullwidth	U+FF01 - U+FF5E (Fullwidth Forms)	Wide monospace characters	Most browsers
Leet Speak	ASCII substitution (A→4, E→3, etc.)	Hacker-style alphanumerics	Universal
Upside Down	Latin Extended / IPA Extensions	Inverted characters, reversed order	Most browsers
Strikethrough	U+0336 (Combining Long Stroke Overlay)	Line through each character	Universal
Encircled	U+20DD (Combining Enclosing Circle)	Circle around each character	Desktop browsers
Squared	U+20DE (Combining Enclosing Square)	Square around each character	Desktop browsers
Bubble Text	U+24B6 - U+24E9 (Enclosed Alphanumerics)	Letters inside circles	Universal
Small Caps	Latin Extended-B / IPA	Reduced uppercase forms	Most browsers
Math Bold	U+1D400 - U+1D433 (Math Alphanumeric)	Bold serif mathematical symbols	Modern browsers
Math Italic	U+1D434 - U+1D467 (Math Alphanumeric)	Italic serif mathematical symbols	Modern browsers
Fraktur	U+1D504 - U+1D537 (Math Alphanumeric)	Gothic/Blackletter script	Modern browsers
Double-Struck	U+1D538 - U+1D56B (Math Alphanumeric)	Outlined "blackboard bold"	Modern browsers
Monospace	U+1D670 - U+1D6A3 (Math Alphanumeric)	Fixed-width serif	Modern browsers
Script/Cursive	U+1D49C - U+1D4CF (Math Alphanumeric)	Flowing handwritten style	Modern browsers
Vaporwave	Fullwidth + space insertion	W i d e s p a c e d text	Universal
Morse Hybrid	Dot/dash insertion between chars	t·e - x·t visual noise	Universal

Frequently Asked Questions

Each operating system uses different text shaping engines and font fallback chains. iOS uses Core Text with the San Francisco font stack, which aggressively clips excessive combining marks beyond approximately 5-7 per base character. Android's HarfBuzz engine with Roboto/Noto is more permissive and will render 15+ stacked marks. Desktop browsers (Chrome, Firefox) using FreeType or DirectWrite typically show the most faithful rendering. Test at your target intensity on the actual device before distributing.

Combining characters are valid Unicode (conformant per Unicode Standard Annex #15). UTF-8 encoded databases (MySQL utf8mb4, PostgreSQL) store them correctly. However, text fields with character limits count combining marks individually - a single Zalgo-corrupted letter at intensity 100 may consume 20+ characters of a 280-character limit. Some WAFs (Web Application Firewalls) flag high combining-mark density as potential homograph attacks per Unicode TR#36.

Homoglyph substitution replaces Latin codepoints with visually identical characters from other Unicode blocks. Latin 'a' (U+0061) becomes Cyrillic 'а' (U+0430). To a human reader, the text looks unchanged. To software performing string comparison, search indexing, or keyword filtering, the text is entirely different. This is the mechanism behind IDN homograph attacks on domain names (documented in RFC 5892). The generator uses this for creative/aesthetic purposes only.

Twitter/X: intensity 30-50 renders well; above 70, tweets may display inconsistently across clients. Discord: supports high intensity (80-100) in desktop but mobile clips aggressively. Instagram: strips most combining marks from bios but preserves them in DMs. Facebook: moderate support, intensity 40-60 recommended. Reddit: full support in comments with old.reddit rendering better than new Reddit for extreme Zalgo.

For Zalgo and strikethrough modes: yes. Stripping all characters in Unicode categories Mn (Nonspacing Mark) and Me (Enclosing Mark) via regex /[\u0300-\u036F\u0489\u1DC0-\u1DFF\u20D0-\u20FF\uFE20-\uFE2F]/g recovers the base text. For homoglyph and leet substitutions: only if you have the exact mapping table used, since the substitution is lossy (multiple source chars may map to the same target). Upside-down text is reversible via the inverse character map.

No. All corruption algorithms execute client-side in your browser using JavaScript string operations on Unicode codepoints. No network requests are made. Your input text is stored only in your browser's LocalStorage for session persistence and is never transmitted externally. You can verify this by monitoring the Network tab in DevTools while using the tool.