Guides › The EPUB format · Published 12 Jun 2026 · Updated 2 Sep 2026 · toolkit.bot

Multilingual EPUB: How to Handle Multiple Languages in One Ebook

June 12, 2026  ·  7 min read

Contents

Bilingual books, language learning materials, annotated translations, and academic texts often mix multiple languages in a single EPUB. Correct language tagging is essential for text-to-speech pronunciation, spell checking, hyphenation, and screen reader behavior. This guide covers how to structure multilingual content in EPUB3.

Why language tagging matters

When a TTS engine or screen reader encounters the sentence 「こんにちは」said John, it needs to know that こんにちは is Japanese (and should use a Japanese voice/pronunciation) while the surrounding text is English. Without language tags, TTS engines either skip non-default-language characters or mispronounce them badly.

Correct language tagging also affects:

Setting the primary language in the OPF

The OPF metadata declares the primary language of the publication. For a bilingual English/French book:

<metadata>
  <dc:language>en</dc:language>
  <dc:language>fr</dc:language>
  ...
</metadata>

The first dc:language element is the primary language. Additional elements declare secondary languages present in the publication. Reading systems use this to set the default TTS voice and hyphenation.

Inline language switches with xml:lang

For content that switches language at the element level, use the xml:lang attribute (and the HTML5 lang attribute — both are valid in XHTML):

Document-level language

<html xmlns="http://www.w3.org/1999/xhtml"
      xmlns:epub="http://www.idpf.org/2007/ops"
      xml:lang="en"
      lang="en">

Paragraph-level switch

<p>The French translation reads:
  <span xml:lang="fr" lang="fr">
    Le renard brun rapide saute par-dessus le chien paresseux.
  </span>
</p>

Section-level switch

<section xml:lang="de" lang="de">
  <h2>Kapitel 1</h2>
  <p>Dies ist der deutsche Text des Kapitels...</p>
</section>

Parallel text (interlinear / facing page)

For language learning books with parallel text:

<div class="parallel-text">
  <p class="source" xml:lang="la" lang="la">
    Gallia est omnis divisa in partes tres.
  </p>
  <p class="translation" xml:lang="en" lang="en">
    All of Gaul is divided into three parts.
  </p>
</div>

Right-to-left content

When mixing left-to-right and right-to-left scripts (e.g., English and Arabic, or English and Hebrew), set the dir attribute alongside xml:lang:

<span xml:lang="ar" lang="ar" dir="rtl">
  مرحبا بالعالم
</span>

For entire chapters in a right-to-left script, set dir="rtl" on the body or section element. In the OPF spine, mark the entire publication's page progression direction:

<spine page-progression-direction="rtl">

This flips the page-turn direction in the reading system (swipe right to go forward). For a bilingual LTR/RTL book, the page progression direction typically follows the primary language. Set the direction in CSS as well, so text alignment and bidirectional punctuation behave:

body {
  direction: rtl;
  text-align: start;   /* logical value — right edge in RTL */
  unicode-bidi: embed;
}

Vertical text and ruby annotations (Japanese and Chinese)

Traditional Japanese and Chinese publishing sets text in vertical columns that run right to left (tategumi). EPUB 3 supports this through CSS writing modes, combined with page-progression-direction="rtl" in the spine so the columns page in the right order. Kobo and Apple Books render vertical text well; Kindle support is partial.

body {
  writing-mode: vertical-rl;
  -webkit-writing-mode: vertical-rl;
}
h1 { writing-mode: horizontal-tb; }   /* horizontal headings inside vertical text */

Pronunciation guides above CJK characters — furigana in Japanese, bopomofo in Chinese — use HTML5 ruby markup, which EPUB 3 supports natively:

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

Font considerations for multilingual EPUB

Embedding fonts in multilingual EPUBs requires care:

FAQ

What BCP 47 codes should I use for language tags?

Use the shortest unambiguous tag: en for English, fr for French, zh-Hans for Simplified Chinese, zh-Hant for Traditional Chinese, pt-BR for Brazilian Portuguese. The full BCP 47 registry is at iana.org/assignments/language-subtag-registry.

Does toolkit.bot preserve language tags when converting PDF to EPUB?

toolkit.bot sets the document language based on the PDF metadata and detected content language, writing both dc:language and xml:lang. RTL documents get page-progression-direction="rtl" automatically, and OCR of scanned non-Latin documents uses the matching language model. For PDFs with mixed languages, the primary language is set in the OPF; inline language switching requires post-conversion editing in Sigil or a text editor.

What is WCAG 3.1.2 (Language of Parts)?

WCAG success criterion 3.1.2 requires that any passage in a language different from the page's primary language be identified with a language tag. This allows assistive technologies to pronounce the text correctly. Required for EPUB Accessibility 1.1 conformance.

How do I create a bilingual facing-page EPUB?

Use a two-column CSS layout or alternate chapters/sections approach. The parallel text markup example above works well for interlinear layouts. For side-by-side pages, fixed-layout EPUB3 gives you precise column control but sacrifices reflowability.

Need a multilingual PDF converted to EPUB?
toolkit.bot detects document language and sets correct OPF metadata — free, no account required.