How do I set the language of an EPUB?

Set dc:language in content.opf metadata using a BCP 47 language tag (e.g., 'ar' for Arabic, 'ja' for Japanese, 'zh-Hans' for Simplified Chinese). Also set xml:lang and lang attributes on the html element of each XHTML content file. Both are needed for full compatibility.

How do I make an EPUB read right to left?

Set page-progression-direction='rtl' on the spine element in content.opf. Add dir='rtl' to the html element in each XHTML file. In CSS, set direction: rtl and text-align: right on the body. The spine direction flips page-turn navigation; the HTML and CSS attributes control text flow within pages.

Does EPUB support Japanese vertical text?

Yes. EPUB 3 supports CSS writing modes for vertical text. Use writing-mode: vertical-rl for traditional Japanese tategumi layout (vertical columns flowing right to left). Also set page-progression-direction='rtl' in the spine. Apple Books and Kobo have the best vertical text support; Kindle support is partial.

How do I handle mixed languages in an EPUB?

Use the lang attribute on inline elements to mark language switches: English phrase inside a French document. This affects hyphenation, font selection, and screen reader language switching. The html element lang sets the default; inline lang overrides it for specific passages.

How do I embed CJK fonts without huge file sizes?

CJK fonts contain tens of thousands of glyphs and can be 5–20 MB. Use pyftsubset to extract only the characters that appear in your document: pyftsubset NotoSansCJK.ttf --text-file=book_chars.txt --output-file=subset.ttf. Generate the character list with: python3 -c "import sys; print(''.join(set(open(sys.argv[1]).read())))" book.txt > book_chars.txt

Multilingual EPUBs: RTL Text, CJK, and Language Attributes

June 12, 2026 · 7 min read

EPUB supports multilingual content, right-to-left scripts, and mixed-language documents — but each requires specific markup. Here's how to handle non-English EPUBs correctly.

Setting the Document Language

The primary language goes in the OPF metadata and in each XHTML file's html element:

<!-- In content.opf -->
<dc:language>ar</dc:language>     <!-- Arabic -->
<dc:language>ja</dc:language>     <!-- Japanese -->
<dc:language>zh-Hans</dc:language> <!-- Simplified Chinese -->
<dc:language>zh-Hant</dc:language> <!-- Traditional Chinese -->

<!-- In each XHTML file -->
<html xmlns="http://www.w3.org/1999/xhtml"
      xml:lang="ar" lang="ar"
      xmlns:epub="http://www.idpf.org/2007/ops">

Use BCP 47 language tags. Both xml:lang and lang are needed — XHTML parsers use xml:lang, HTML parsers use lang.

Right-to-Left (RTL) Text — Arabic, Hebrew, Persian

RTL content requires the reading direction set at both the EPUB spine level and the CSS level:

<!-- In content.opf spine -->
<spine page-progression-direction="rtl">
  <itemref idref="ch1"/>
</spine>

<!-- In the XHTML html element -->
<html xml:lang="ar" lang="ar" dir="rtl">

<!-- In CSS -->
body {
  direction: rtl;
  text-align: right;  /* or text-align: start; for logical properties */
  unicode-bidi: embed;
}

The page-progression-direction="rtl" in the spine tells reading systems to flip the page-turn direction (swipe left to go forward becomes swipe right).

Vertical Text — Japanese and Chinese

Traditional Japanese and Chinese publishing uses vertical right-to-left text (tategumi). EPUB 3 supports this via CSS writing modes:

body {
  writing-mode: vertical-rl;  /* vertical, right to left columns */
  -webkit-writing-mode: vertical-rl;
}

/* For horizontal headers within vertical text */
h1 { writing-mode: horizontal-tb; }

Set page-progression-direction="rtl" in the spine for vertical Japanese (columns flow right to left). Kobo and Apple Books have the best vertical text support; Kindle support is partial.

Mixed-Language Content

For inline language switches (e.g., English quotes in a French document), use lang on the specific element:

<p lang="fr">
  Le terme anglais <span lang="en">"reflowable"</span> n'a pas
  d'équivalent direct en français.
</p>

This affects hyphenation, text rendering, and screen reader pronunciation. Screen readers switch to the correct voice/language when they encounter a lang attribute change.

Ruby Annotations (CJK Pronunciation)

Ruby annotations show pronunciation guides above CJK characters (furigana in Japanese, bopomofo in Chinese):

<ruby>漢<rt>かん</rt></ruby><ruby>字<rt>じ</rt></ruby>

EPUB 3 supports HTML5 ruby markup natively. Apple Books renders ruby well; Kindle support varies by device generation.

Font Considerations for Non-Latin Scripts

Arabic/Hebrew — most system fonts include Arabic and Hebrew; embed only if using a specific typeface. Noto Sans Arabic is freely embeddable (SIL OFL license).
CJK — CJK fonts are very large (5–20 MB). Subset to only the characters used with pyftsubset: pyftsubset NotoSansCJK.ttf --text-file=book_chars.txt
Fallback — always specify a generic fallback: font-family: "MyFont", serif;

Converting Non-English PDFs to EPUB

toolkit.bot detects the document language from PDF metadata and sets the correct dc:language and xml:lang attributes in the output EPUB. RTL documents get page-progression-direction="rtl" automatically. OCR for scanned non-Latin documents uses the appropriate language model.

Convert multilingual PDF to EPUB →