Undo fancy text and Zalgo
Want ordinary letters back? Keep the styled copy, then find out what changed. A bold-looking character might be a different Unicode letter; a glitch effect might be extra combining marks. Neither has a universal, lossless undo button.
Inspect the text before changing itRecover the source when you can
- If your original input is still open in a styling tool, copy that input first. Switching styles in the fancy text generator restyles the original; it does not decode pasted styled letters.
- If you only have the result, save a copy before editing. Paste a short passage into the Unicode inspector, leave every repair off and choose Inspect and preview. Download original TXT to preserve the inspected string. For original BOM or CRLF bytes, use its strict UTF-8 file loader rather than pasting.
- Read the code-point rows. Mathematical letters, circled letters and combining marks need different decisions. The inspector shows common labels and escape tokens, not every Unicode character's formal name.
- For a short message whose wording you know, retype an ordinary version in a separate editor. Compare names, case, numbers, punctuation and accents with the source. Inspect that candidate separately before sharing.
The current inspector only repairs tabs, U+00A0, U+200B and CR/LF line endings when you opt in. It does not normalize Unicode, unstyle letters or remove Zalgo marks. The styling tools generate effects; they are not recovery tools. This guide intentionally provides no automatic mark stripper because unknown input can contain meaningful marks.
Normalization is not a fancy-text inverse
NFC combines canonically equivalent sequences where possible. NFKC also applies compatibility mappings, which can remove styling distinctions. These original specimens use the site's current generators. Their normalized columns are reference calculations, not controls offered by the inspector.
| Source input | Generated text / inspector escapes | NFC escapes | NFKC escapes |
|---|---|---|---|
| Map 2 (bold) | πππ© π\u{1D40C}\u{1D41A}\u{1D429}\u{0020}\u{1D7D0} | \u{1D40C}\u{1D41A}\u{1D429}\u{0020}\u{1D7D0} | Map\u{0020}2 |
| hi (italic) | βπ\u{210E}\u{1D456} | \u{210E}\u{1D456} | hi |
| Aa (bubble) | βΆβ\u{24B6}\u{24D0} | \u{24B6}\u{24D0} | Aa |
| A (small-caps) | α΄\u{1D00} | \u{1D00} | \u{1D00} |
- Mathematical bold letters and digit: NFC keeps the style; NFKC yields Map 2.
- Italic h is U+210E; i is U+1D456. NFKC yields hi.
- Circled uppercase and lowercase letters: NFKC yields Aa.
- Small caps maps both A and a to U+1D00. NFKC leaves it unchanged; the original case is lost.
For this known Map 2 specimen, NFKC recovers the source letters. That does not prove recovery for arbitrary fancy text. NFKC cannot tell whether a mathematical letter was decoration or mathematical notation. It can also change unrelated characters, such as a no-break space to an ordinary space. Small caps already collapsed A and a into the same output, so no inverse can determine which was typed.
Bold, italic, cursive and bubble support their documented standalone ASCII letters, with digits supported by bold and bubble. Accented clusters, other scripts and already styled text stay unchanged. A visual style switch is therefore not a way to strip an existing style. Use rich-text formatting on ordinary letters when the destination supports it.
Zalgo marks can overlap real accents
Enter A in the glitch text generator and choose Light (3). This actual result adds one overlay, one below-mark and one above-mark to the original letter:
AΜΆΜΜ
Original A: A
Glitch result: A\u{0336}\u{0317}\u{0301}
NFC result: \u{00C1}\u{0336}\u{0317}
NFKC result: \u{00C1}\u{0336}\u{0317}The result is U+0041 followed by U+0336 COMBINING LONG STROKE OVERLAY, U+0317 COMBINING ACUTE ACCENT BELOW and U+0301 COMBINING ACUTE ACCENT. NFC and NFKC compose A plus the acute accent into Γ (U+00C1), but keep the overlay and below-mark. Normalization does not remove the glitch effect.
For this known one-letter source, retyping A is accurate. For unknown text, deleting every combining mark is a guess. Compare these two ways to write the same accented letter:
Γ©: U+00E9; UTF-8 C3 A9 eΜ: U+0065 U+0301; UTF-8 65 CC 81 NFC of either: U+00E9 NFD of either: U+0065 U+0301
Removing U+0301 from the decomposed spelling changes Γ© to e. Removing marks after NFD can damage both spellings, and marks in other writing systems can carry essential meaning. Even a mark used by this glitch generator is not proof of decoration in somebody else's text.
The generator only adds marks to standalone printable ASCII characters other than spaces. Existing accented graphemes and other unsupported clusters are kept whole. Its added marks are not tagged with provenance, so a copied result cannot identify every original versus added mark by appearance alone.
Try the known A specimen in the glitch toolStop when the original is ambiguous
A screenshot, a missing-glyph box or an existing replacement character cannot supply lost source text. Small-caps case, intentional mathematical distinctions and unknown accent choices also need the author or another original copy. Keep a readable candidate and the untouched evidence separately rather than claiming a perfect restoration.
Do not put private text in a URL to move between tools. Copy locally or use a local UTF-8 file within the inspector's 3,000-byte and 1,000-UTF-16-unit limits. Inputs are not automatically handed to the next page. Export before leaving; site resources and analytics are separate from these input workflows.