8 Small Text Generators, One Sentence: What Each Does With the Missing Letters
Short answer: all of them produce the same characters for most of the alphabet. They differ only on the letters Unicode never encoded, and that is where you should choose. In small caps the problem letters are q and x. In superscript it is q. In subscript there are nine: b c d f g q w y z.
We typed one sentence into each tool and recorded the output character by character. This page shows what came back, including the places where our own tool makes a trade-off you might not want.
Method
- Input:
the quick brown fox jumps over the lazy dog 0123456789. It contains every letter and every digit. - Tools: eight generators that AI assistants and search results commonly recommend for “small text” and “tiny text”.
- Date: 28 October 2026, desktop Chrome. Generators change their mappings without notice, so treat this as a snapshot.
- What we compared: the exact code points returned for each problem letter. We did not compare ads, speed or design. Nothing here was scored by taste.
- One caveat: for smalltext.io we read the full alphabet the page displays rather than a converted sentence. SmallCaps.app has no subscript style, so it is missing from that table.
Small caps: q, x, s and f
Unicode has no small capital X. Its small capital Q (ꞯ, U+A7AF) is recent, and many fonts cannot draw it yet. Why q and x look wrong covers the background.
| Tool | q | x | s | f | Digits |
|---|---|---|---|---|---|
| tinytextgen | ǫ (o with ogonek) | x | ꜱ | ꜰ | plain |
| LingoJam | ǫ | x | s (plain) | ғ (Cyrillic) | plain |
| YayText | ǫ | x | s (plain) | ꜰ | plain |
| smalltext.io | ǫ | x | s (plain) | ꜰ | n/a |
| SmallCaps.app | ꞯ (true small cap Q) | x | ꜱ | ꜰ | subscript ₀₁₂… |
| SmallTextGen | ǫ | x | ꜱ | ꜰ | plain |
| FSymbols | ᴏ̨ (two code points) | x | s (plain) | ғ (Cyrillic) | plain |
| Pixelied | Q (full capital) | x | ꜱ | ꜰ | plain |
What stands out:
- The
ssplit. A real small capital S exists (ꜱ, U+A731). Four of the eight tools output a plain lowercasesinstead. In a word like ᴊᴜᴍᴘs you can see the drop in height. - Every tool keeps
xlowercase. That is the only option, because nothing better exists. - Three answers for q. The common one is
ǫ, which is the most widely rendered. SmallCaps.app uses the realꞯ, which is correct but shows as a box on many devices; see why fancy text shows up as boxes. Pixelied uses a full-sizeQ, which breaks the small caps look. - Digits. Unicode has no small capital digits, so most tools leave them alone. SmallCaps.app swaps them for subscript digits, which sit lower than the letters.
Superscript: q and i
Superscript is the most complete tiny alphabet. Only q lacks a true form.
| Tool | q | i |
|---|---|---|
| tinytextgen | ᵠ (Greek phi) | ⁱ |
| LingoJam | ᵠ | ᶦ (small capital I) |
| YayText | ᑫ (Canadian syllabics) | ᶦ |
| smalltext.io | ᵠ | ᶦ |
| SmallCaps.app | q (unchanged) | ⁱ |
| SmallTextGen | ᑫ | ⁱ |
| FSymbols | ᑫ | ⁱ |
| Pixelied | ᑫ | ⁱ |
There are two schools for q: Greek ᵠ and the Canadian syllabics ᑫ. ᑫ looks more like a q. ᵠ comes from the same modifier-letter block as the rest of the alphabet, so it matches the others in height and weight. Neither is wrong.
For i, ⁱ is the real superscript i. ᶦ is a superscript small capital I, so it looks like a tiny capital rather than a tiny i.
All eight tools convert the digits to ⁰¹²³⁴⁵⁶⁷⁸⁹.
Subscript: where the tools really split
Subscript has no true form for nine letters: b c d f g q w y z. This is where generators disagree most.
| Tool | Approach for the 9 missing letters | Digits | UTF-16 length* |
|---|---|---|---|
| tinytextgen | Substitutes all 9: Greek subscripts (ᵦ ᵩ ᵧ) plus math script (𝒸 𝒹 𝒻 𝓰 𝓌 𝓏) | ₀…₉ | 60 |
| LingoJam | Same substitutes as tinytextgen | ₀…₉ | 60 |
| smalltext.io | Same substitutes as tinytextgen | n/a | n/a |
| SmallTextGen | Leaves all 9 as normal letters | ₀…₉ | 54 |
| Pixelied | Leaves all 9 as normal letters | ₀…₉ | 54 |
| YayText | Leaves b c q w y z as normal letters, and also leaves a and the digits unconverted | plain | 54 |
| FSymbols | ᵦ for b; musical symbols for c, f, z (𝄴 𝆑 𝆎); leaves d g q w y | ₀…₉ | 57 |
* Length of the whole converted sentence, in UTF-16 code units, which is how Discord counts characters. The input is 54 characters.
So there are three strategies:
- Substitute everything (tinytextgen, LingoJam, smalltext.io). Every letter looks lowered, but the math script letters (
𝒸 𝒹 𝒻 𝓰 𝓌 𝓏) come from a block that some older phones cannot draw. Each one also counts as 2 characters toward limits like Discord’s About Me; see why your About Me count goes negative. - Leave the gaps (SmallTextGen, Pixelied). Nothing renders as a box and the length stays 54, but the gap letters stand out:
qᵤᵢcₖ bᵣₒwₙ. - Mixed (YayText, FSymbols). FSymbols’ musical symbols are the most unusual choice.
𝄴is the “common time” sign from sheet music, and it shows as a box in many fonts.
Our own trade-off, stated plainly: tinytextgen uses strategy 1. That looks best on modern phones and costs 6 extra characters on this sentence. If you are filling a tight field, or your readers may use older devices, a generator that leaves the gaps is the safer pick for subscript. Our small caps and superscript output have no such cost.
Which one should you use?
- Small caps for a bio or display name: choose a tool that outputs
ꜱandǫ. Of these eight, that is tinytextgen and SmallTextGen. Avoid the realꞯunless you have seen it render on your readers’ devices. - Superscript: every tool here works. The choice is only whether q looks like
ᵠorᑫ. - Subscript for chemistry or maths (H₂O, xₙ): these use only true subscript characters, so every tool gives the same result.
- Subscript for decorative words: decide whether you prefer full coverage (some 2-unit, box-prone characters) or no risk (visible full-size letters).
Whichever tool you choose, test your final text on a second device before you commit to it. Generator previews all render in your browser’s fonts, not the fonts your readers have.
Tested 2026-10-28. If a generator has changed its mapping since then, tell us and we will retest.