Directionality Studio
Scripts do not all run the same way. Four directions, four living examples.
Greek, Latin, Cyrillic, and most of the world
Arabic, Hebrew, N'Ko, Adlam, Thaana
Mongolian — columns run left to right
ΑΛΦΑΒΗΤΟΝ
ΑΛΦΑΒΗΤΟΝ
How early Greek ran — "as the ox plows". Rongorongo used a reverse form (see the Undeciphered Vault).
Anatomy Studio
The words a designer uses for the parts of a letter — baseline, x-height, cap height, ascender, descender — were built for one script and describe it well. Lay the same five lines over another and watch how many of them have nothing to sit on. The vocabulary is not wrong; it is local.
The vocabulary fits
All five hold. H reaches cap height, b goes higher still to the ascender, x defines the x-height, and p and g drop below the baseline. This is the script the vocabulary was written for.
All five hold here too, and not by coincidence — the Latin alphabet took its letterforms from Greek. β reaches above and below in one stroke, which no Latin letter does.
The five exist, and the lowercase barely uses them. б is close to the only ascender in the alphabet, and most letters sit entirely between baseline and x-height — which is why a Cyrillic paragraph reads flatter and more even than a Latin one.
It half fits
A baseline, and letters that fall well below it. But there is no capital form, so cap height has nothing to measure, and the x-height rule is an arbitrary line across a cursive join.
The letters hang from a headline — the shirorekha — rather than standing on a baseline. The strongest horizontal in the word is at the top, which is the one place the Latin vocabulary has no word for.
It does not apply
Every character fills the same em square. There is no baseline to sit on and no ascender to reach — the two rules that matter are the top and bottom of the box, and all five guides describe nothing.
The same square, filled differently. The strokes curve and rarely reach the edges, so the weight sits inward and a line of hiragana looks lighter than a line of Han at the identical size.
The same square again, and the same size — but the strokes are straight and run to the edges. Set hiragana and katakana side by side and the difference is density, not height.
Square blocks like Han, but each one is built from letters arranged inside it — jamo, assembled into a syllable. There is a level of structure between the stroke and the syllable, and the Latin vocabulary has no word for that either.
Five lines, nine scripts, one vocabulary. Toggle the guides to see which survive. The dashed red rules are the ones that have nothing to measure — they are drawn to make the absence visible, not to mark a deficiency in the script.
Measure Studio
The same words, the same face, the same size. The only thing that changes is how far the eye travels before it turns back. Read the left column, then the right, and notice where you lose your place.
45 characters — a comfortable measure
130 characters — the same paragraph, far wider
This screen is not wide enough to set the second paragraph any wider than the first, so the two above are currently the same measure. The slider below still works.
The figure typographers usually give is 45–75 characters for continuous reading; the Latin tradition settles near 66. It is a property of the line, not the typeface — which is why a wider column cannot be fixed by changing the font, and why this studio holds the font still.
The Tofu Room
When a character has no glyph to show, you get □ — a blank box the type world calls "tofu". This whole building is set in Noto, whose name is a contraction of "no tofu".
What your device shows for a character it cannot render.
Two things cause tofu. First, the character is simply not encoded in Unicode at all — there is no code point, so there is nothing for any font to render. This is the case for the Proto-Sinaitic ancestor and the Indus signs: the marks exist on stone, but not yet in the standard.
Second, the character is encoded, but the reader's device lacks a font that covers it — the code point is valid, yet no glyph is installed to draw it. This is exactly the problem Noto was built to end: one type family spanning every script in Unicode. If you see □ anywhere in these halls, your device is missing a Noto font — the very problem on display here. The unencoded cases live down in the Undeciphered Vault.
Specimen Press
Set any script in the collection live. Choose a script, drag the size, and type — every specimen renders in its own Noto face.
Script —
The Keyboard Room
How do you actually type a script? Not all the same way. Pick a layout to see its real, standard keyboard — the Latin legend in the corner of each key, the script's own character on the cap. The three letter rows are shown; toggle the Shift layer to see how a script with more signs than keys overflows.
Three ways a script meets a keyboard. Direct — one key per sign, Latin and script sharing each cap (Russian ЙЦУКЕН, Greek, Arabic, Hebrew). Layered — more signs than keys, so Shift and the number row carry the overflow (Cherokee's 85 syllables). Composition — far too many signs for any keyboard: Chinese is typed in Pinyin and chosen from a candidate list; Japanese is typed as romaji or kana and converted; Korean types single jamo that the system assembles live into syllable blocks — the bridge between a keyboard and an input method.
Letter-row layouts transcribed from the shipping keyboard drivers via kbdlayout.info and standards documentation. Cherokee is the Cherokee Nation layout (positional, not phonetic); its number-row syllables are omitted here.