Tiny Text Generator

Tiny text is not one thing. It is three separate sets of Unicode characters that were encoded for phonetics and mathematics decades ago and never intended as a styling option, which is why two of them are missing letters and one of them is not. Getting a good result is mostly a matter of knowing which set you are using and what it leaves out.

The Font Generator is a free Unicode text tool with 35 styles and 116 decorations, made for anyone styling a social profile or gaming name, and it runs entirely in your browser with no sign-up.

What is a tiny text generator?

A tiny text generator replaces each letter you type with a Unicode character that was drawn small. It is not a font size setting and it changes nothing about the text box: the letters themselves are different letters, which is why the small look survives being copied into an app that has no formatting at all.

Three sets do the work. Small capitals come mostly from IPA Extensions and Phonetic Extensions, where linguists needed a reduced capital to mark a particular sound. Superscript letters are modifier letters from the same phonetic tradition. Subscript letters come from the Superscripts and Subscripts block, encoded so that chemistry and mathematics could write H₂O and x₁ as text.

None of the three was designed to spell words, and that single fact explains every gap, every substitution and every reverted bio on the rest of this page.

What is the difference between small caps, superscript and subscript?

Small caps keep every letter on the baseline at roughly capital-height-reduced size. Superscript raises the letters toward the cap line. Subscript drops them below the baseline. They look related on screen and behave completely differently once a platform starts processing them.

The column that matters most is the last one. If you want a small look that stays put in a field you do not control, small caps is the only one of the three that is not flagged as a styled variant of something else.

Which tiny letters actually exist in Unicode?

Here is the full a to z of each set as this tool emits it. Anything printed at full size in these rows is a letter that has no small form to swap in, left as an ordinary character so the word still reads.

Some generators hide their gaps by borrowing from other alphabets, such as a Cyrillic ғ where a small capital F should be, or a Greek phi standing in for a raised q. The characters look approximately right and belong to a different script entirely, which is how a name ends up failing a validation check for reasons nobody can see. This tool leaves the gap as a plain letter instead, so what you copy is what you typed.

Why is there no superscript q?

Because nobody ever needed one. The raised Latin letters were encoded to serve phonetic transcription and mathematical notation, and both fields proposed the characters their own conventions required. No convention ever called for a raised q, so it was never proposed and never encoded. Twenty-five of the twenty-six exist; q is the one that does not.

You can see the ad hoc history in where they live. Raised h, j, l, r, s, w, x and y sit in Spacing Modifier Letters at U+02B0 and above. Raised a, b, d, e, g, k, m, o, p, t, u and v are in Phonetic Extensions from U+1D2C. Raised c, f and z are in Phonetic Extensions Supplement from U+1D80. Only ⁱ at U+2071 and ⁿ at U+207F are in the Superscripts and Subscripts block that shares their name, alongside all ten digits.

Three of the digits are stranded somewhere else again: ¹, ² and ³ are at U+00B9, U+00B2 and U+00B3 in Latin-1 Supplement, encoded early for compatibility with legacy character sets, while ⁰ and ⁴ to ⁹ came later at U+2070 and U+2074 onward.

Which letters are missing from subscript?

Nine of them: b, c, d, f, g, q, w, y and z. Subscript has 17 letters in total (a, e, h, i, j, k, l, m, n, o, p, r, s, t, u, v and x), and that is the complete set Unicode defines. It is the least complete alphabet on this page by a wide margin.

The reason is the same as the superscript q, only more so. Subscript letters exist for variable indices and chemical notation, and the handful of letters those uses needed is the handful that got encoded. Twelve of them are in Superscripts and Subscripts from U+2090, four more are scattered through Phonetic Extensions, and ⱼ was picked up separately at U+2C7C.

In practice this rules subscript out for most names. Try it on anything containing a b, a d, a g or a y and the word comes back half-dropped and half-lowered, which reads as a mistake rather than a style. Subscript is genuinely good for digits and for short words that happen to fall inside the seventeen; for anything else, small caps is the set that can spell it.

Does tiny text use up your character limit?

No, and this is the one place where tiny text beats every other style on the site. Every small cap, raised letter and lowered letter sits below U+FFFF, so each one is a single UTF-16 unit, exactly the same cost as the plain letter it replaces.

Bold and cursive come from Mathematical Alphanumeric Symbols above U+FFFF and are stored as surrogate pairs, so they cost two units per letter. A 150 character Instagram bio holds 150 small caps and about 75 cursive letters. On an 80 character TikTok bio or a 32 character Discord display name, that difference decides whether the line fits.

Why does your tiny text turn back into normal letters?

Because superscript and subscript characters are formally defined as styled versions of ordinary letters. ᵗ carries a compatibility decomposition marked <super> pointing at t, and ₜ carries one marked <sub> pointing at the same letter. Those mappings are part of the standard, not a quirk of any one app.

When a field applies NFKC normalisation, the compatibility form set out in Unicode Standard Annex #15, it follows every one of those decompositions and rebuilds the string from the plain letters. ᵗⁱⁿʸ goes in and tiny comes out. Username fields normalise on purpose to stop two accounts looking identical, and search indexes normalise so that a query matches text regardless of how it was styled.

Small caps are the exception, and it is a useful one. ᴀ at U+1D00, ʙ at U+0299, ᴇ at U+1D07 and ᴋ at U+1D0B are encoded as independent letters with no decomposition at all, so NFKC leaves them exactly as they are. That gives you a rule worth remembering: if a field keeps flattening your tiny text, stop fighting it and switch to small caps.

How do you make small text in Discord?

Inside a message, use Discord’s own subtext instead of pasting characters. Type a hyphen, a hash and a space at the very start of the line and Discord renders the whole line smaller and greyed out. That is real small text: it stays searchable, it reads correctly to assistive software, and it does not depend on which characters exist.

1. In a message box, start the line with -# followed by a space, then type normally.
2. Keep -# as the first thing on the line, because a space or any other character before it and Discord treats the line as ordinary text.
3. For a nickname, server name, channel name or custom status, paste small caps from the generator above instead, because those fields do not parse markdown.
4. Check the result in the member list, where Discord draws names at their smallest.

Between the two, subtext handles anything you are saying and pasted characters handle anything you are called. Discord fonts covers which of Discord’s name fields accept styled Unicode and which one refuses it outright.

Is Unicode subscript the same as subscript in Word or Google Docs?

No, and mixing them up causes real problems in documents. Word, Google Docs, Excel and PowerPoint apply subscript as character formatting: the letter underneath stays an ordinary letter and the application simply draws it lower and smaller. Unicode subscript replaces the letter with a different character.

The consequences diverge quickly. Formatted subscript can be searched with the base letter, spell-checked, turned off with the same keyboard shortcut that turned it on, and it covers all 26 letters because nothing is being looked up. Unicode subscript survives copy and paste into apps that have no formatting at all, and pays for it by being unsearchable, unremovable and incomplete.

Use the formatting in a document. Use the characters where there is no formatting to use: a bio, a display name, a comment box, a game tag, a spreadsheet cell whose value must stay a single string.

Why does tiny text look uneven?

Because a single line of it is assembled from several Unicode blocks that were designed separately. As the earlier breakdown shows, a superscript word can pull characters from Spacing Modifier Letters, Phonetic Extensions, Phonetic Extensions Supplement and the Superscripts and Subscripts block all at once.

Those blocks were drawn at different times for different purposes, so their x-heights, stroke weights and side bearings were never reconciled. ᵗⁱⁿʸ ᵗᵉˣᵗ therefore sits slightly ragged rather than looking like one small typeface, and a device that has to pull glyphs from more than one font to cover the line makes the mismatch more obvious again.

Small caps hold together best, because almost all of the letters come from the same two phonetic blocks and were drawn as a set. That is the third reason it tends to be the right default.

What does tiny text sound like to a screen reader?

Mostly like a string of character descriptions, and sometimes like silence. Modifier letters in particular are announced by name where they are announced at all, so a bio written in superscript is closer to unreadable than to small.

There is a second cost that gets less attention. Tiny text is often used to slip a word past keyword filters, and platforms increasingly normalise text before moderating it, the same NFKC pass described above. That means the styling disappears and the word is matched anyway, so you get the accessibility penalty without the effect you were after.

Which tiny text style should you use?

The samples below are rendered from the live generator at build time. The word is quiet on purpose: q is the letter with no superscript form and no small-cap form in this tool, so you can see exactly how a gap looks before you paste one.

1. qᴜɪᴇᴛSmall CapsThe default. Survives normalisation and costs one unit per letter.
2. qᵘⁱᵉᵗSuperscriptRaised. The q stays full size, because no raised q exists.
3. qᵤᵢₑₜSubscriptLowered. Only 17 letters exist, so most words come out mixed.
4. ⋆⁺₊✧ qᴜɪᴇᴛ ✧₊⁺⋆Small Caps + ✧Each decoration character costs a unit, same as a letter.
5. ⟡ qᴜɪᴇᴛ ⟡Small Caps + ⟡A quieter pair of brackets for a bio heading line.

If you need a small look on a word containing q, s or x, read the sample first. Those three come back as ordinary lowercase in small caps, which is fine mid-word and obvious at the start of a name.

When in doubt, use small caps

Three sets, one recommendation. Small caps has a usable alphabet, costs one unit per letter, holds together visually because its characters were drawn as a group, and is the only one of the three that a normalising field leaves alone. Superscript and subscript are better thought of as notation you can borrow for a footnote mark or a formula than as a way to write words.

Inside a Discord message, skip all of it and type -# instead. Everywhere a field stores exactly what you typed, scroll back up, enter your line, and compare the three sets before you commit one to a profile.