Fancy text generator

What this fancy text generator actually generates

The result is text made from Unicode characters. It is not a new font, a font file, or a visual effect applied by CSS. When ordinary A becomes mathematical bold 𝐀, the browser is holding a different code point. That distinction is why the result can be copied into another app without taking a stylesheet along, and it is also why the appearance can change after you paste it. The receiving app chooses whichever installed font can draw that character.

Similar tools are sometimes called font generators, but a real font generator or font editor works with outlines, metrics, kerning, OpenType layout tables and files such as TTF, OTF, WOFF or WOFF2. Nothing on this page opens, creates, installs or modifies those files. It performs a fixed character-to-character mapping in your browser. If you need the vocabulary behind characters, code points and glyphs, what Unicode is explains why the character and the drawing are separate things.

Why the styles have official names

The thirteen mathematical styles come from the Mathematical Alphanumeric Symbols and related Letterlike Symbols assigned by Unicode. They include bold, italic, bold italic, script, bold script, Fraktur, double-struck, bold Fraktur, four sans-serif variants and monospace. The other two cards use Unicode's fullwidth and circled alphanumeric characters. Names such as “cursive” may be popular in search boxes, but the card says “Mathematical Script” because that is the encoded set it actually uses.

This matters at the holes. Several familiar mathematical letters were encoded years before the large mathematical block, so script capital B is U+212C and double-struck capital C is U+2102 rather than the value a simple offset formula predicts. Circled zero is separated from circled one through nine as well. The tables here list every mapping explicitly, including those exceptions, instead of calculating a range and quietly emitting the wrong characters at the gaps.

Latin, Greek and characters that stay unchanged

Every card follows its real Unicode source domain. All mathematical styles cover Latin letters, while only styles with encoded mathematical digits change digits. Five styles also contain the full mathematical Greek set: bold, italic, bold italic, sans-serif bold and sans-serif bold italic. Bold includes digamma, and italic includes the dotless i and j forms assigned by Unicode. A Greek letter in another card stays Greek and stays unchanged; it is never replaced with a Latin or Cyrillic character that merely looks similar.

The same preservation rule covers Arabic, CJK, emoji, accented letters that have no assigned styled counterpart, combining marks, joiners and every other unsupported code point. Nothing is normalized, deleted, reordered or transliterated. Each result reports how many iterator positions changed and how many were preserved, and its receipt lists the exact displayed output code points. For a clearly marked large-input preview, the receipt is limited to that preview while Copy still uses the full result. That lets you distinguish a style that transformed the whole input from one that only changed a few Latin letters around otherwise untouched text.

Copy and paste limits

Copying writes the exact result represented by one card. For ordinary input, that is the complete text shown. Very large input uses a clearly marked, bounded preview while Copy still writes the complete result after local chunked processing. It does not copy HTML, a hidden font-family declaration or an image. That makes the result portable as text, but not universally accepted. A username field may reject compatibility characters, a search system may treat them differently from ordinary letters, and a moderation or spam filter may deliberately normalize them. Some destinations count one mathematical character as two UTF-16 code units even though it is one Unicode code point.

Searchability deserves special caution. Mathematical bold 𝐀 and ordinary A are related by a compatibility decomposition, but software is not required to apply that decomposition to every search, mention, hashtag or account lookup. Do not use styled characters in information that must be reliably found, matched or entered by someone else: email addresses, URLs, legal names, passwords, product identifiers and recovery codes are poor places for decorative Unicode.

Accessibility and platform rendering

Screen readers do not all announce these characters the same way. One may read a mathematical letter by its formal name, another may read the base letter, and another may skip an unsupported character. A long paragraph converted into compatibility characters can therefore become tiring or unintelligible even when it looks polished on your screen. Keep essential wording in ordinary text, especially headings, links, instructions and calls to action. Decorative text should supplement a readable label, not replace it.

Glyph coverage is a separate boundary. Modern operating systems cover most of these sets, but the exact weight, slant, stroke shape, spacing, enclosure and baseline come from local fallback fonts. A card can look balanced in one browser and uneven in a messaging app on another phone while carrying byte-for-byte identical code points. The receipt proves identity, not visual sameness. Test the final text in the actual destination before publishing it.

Private, deterministic character mapping

Typing and pasting stay in this browser tab. There is no upload step, server request, account or model call. Each input code point is checked against a version-pinned Unicode 17 table, and the same text produces the same output every time. Running an already styled result through the same card again leaves it unchanged, because styled targets are not fed back into an offset calculation. Clearing the field clears the results immediately; CreatorValet never receives the text.