Why does my ebook show boxes instead of letters?
The font your reading app draws with has no shape (a glyph) for those characters, and the app found no other font on the device that has one. The text in the file is usually intact. Type professionals call the small empty rectangles tofu, as NPR's Code Switch blog explained in August 2014. The fix is a font that covers the script: the one the publisher embedded in the book, if there is one, or any font on the device that has those characters.
Numbers for characters, shapes from fonts
The first version of the Unicode Standard came out in October 1991. Unicode gives every character a number, its code point, and leaves the look of it to fonts. A file can therefore hold perfectly good Greek or Tamil that a given font can't draw. Even Amazon's own store hedges on this: as of November 2026, Amazon's KDP page on Indian languages lists Hindi, Tamil, Malayalam, Marathi and Gujarati for Kindle ebooks, but recommends typing the title and author name in English, because these languages may not display correctly on every Amazon and Kindle website, app and email.
When a font is asked for a character it lacks, the software falls back on glyph 0. The OpenType recommendations say glyph 0 must be the .notdef glyph, and suggest drawing it as an empty rectangle, a rectangle with a question mark inside, or one with an X. They warn against leaving it blank, because a blank .notdef looks like a space and you'd never know anything was missing. Since every font draws its own .notdef, the boxes in one app can look different from those in another.
Some fonts draw a box that tells you more. The Last Resort font, drawn by Apple, Michael Everson of Evertype and the Unicode Consortium, carries a copyright notice that starts in 1998. Each of its glyphs stands for a whole Unicode block: a sample character of the block inside a square frame, with the block's name and code point range written along the frame, legible at large sizes. A frame around a small Devanagari letter tells you that the passage is in Devanagari, not that the file is damaged.
Noto, named after the boxes
Google released its first Noto fonts in 2012, NPR reported, and the name is short for "no tofu". By August 2014 the family covered 100 scripts and 100,000 characters, which Jungshik Shin, an engineer on Google's text and font team, said took in more than 600 written languages. The project's FAQ still gives the aim as "no more tofu".
The same FAQ explains why Noto comes as many files and not one: it has far more glyphs than a single font can hold, and the Chinese, Japanese and Korean set on its own is already as big as a font can be. The ceiling is in the file format. OpenType keeps the glyph count in a 16-bit field (numGlyphs in the maxp table), so a font can't have more than 65,535 glyphs. Unicode 18.0, released on 16 September 2026, defines 172,808 characters.
No single font can cover Unicode, so apps try one font after another. Ebook apps lay out pages with CSS, and CSS Fonts Level 4 spells out the order: each font the text names, then the fonts installed on the device, and only then a missing glyph symbol, from a Last Resort font or the like. One kind of character never reaches the device fonts. For a code point in Unicode's Private Use Area, the app must look only in the fonts the text names. A publisher who put a decorative initial or a symbol there is betting on their embedded font; if it doesn't load, that character is a box every time.
Where the boxes come from in an ebook
A book doesn't have to rely on the fonts of the device, because EPUB lets it bring its own. The EPUB 3.3 rules for reading systems say they must support TrueType, OpenType, WOFF and WOFF2 fonts that the book's style sheet loads with @font-face. When the publisher embedded a font with the right script and the app uses it, the text displays as the publisher saw it. Boxes come when the book embeds no such font and the device has none, when your settings replace the book's fonts, or when an embedded font fails to load.
Your own font setting is the easiest of these to check. On a Kobo, the Font Face list (tap the middle of the page, then the Text icon) swaps in a font of your choice, and Kobo's help page says the eReader remembers that choice and applies it to the other books you read. A font picked for an English novel may have no Greek or Cyrillic at all. Kobo eReaders accept extra .otf and .ttf files in a folder called fonts on the device, Kobo says, so you can add a font with wider coverage (see adding fonts to a Kobo or a Kindle and why an ebook ignores your font setting).
An embedded font can also fail to load. EPUB 3.3 lets a publisher obfuscate an embedded font so that it can't be installed or reused straight from the book, but the reading system rules say only that apps should undo the obfuscation, not that they must. They also note that EPUB 3 allowed obfuscation before version 3.0.1 without fixing whether it came before or after compression, so an app can end up with an invalid font from an older file. The text then falls back to the device's fonts, which may lack the script.