ePub3 Media Overlays: Synchronized Audio and Text for Read-Along Books
ePub3's media overlays let books read themselves while highlighting text. Here's how to author them, validate them, and deliver to Apple Books and Google Play Books.
ConvertIntoMP4 Team·May 8, 2026·7 min read
What Read-Along Books Actually Are
Children's picture books on iPad. Language-learning materials with synced audio. Accessibility-focused books that read themselves while highlighting current text. These are all built using ePub3's media overlays specification (officially "Media Overlays 3.0").
A media overlay is a SMIL (Synchronized Multimedia Integration Language) file that pairs each text element in the book with a timestamp in an audio file. When the audio plays, the corresponding text highlights. When the user taps text, audio jumps to that point.
The format has been around since 2011 but production tools are rare. This post covers the authoring workflow, validation, and platform-specific quirks for Apple Books, Google Play Books, and Adobe Digital Editions.
For converting other ebook formats, see our ebook converter.
Each <par> (parallel) element pairs a text element (identified by ID in the XHTML) with an audio segment (start and end times).
The XHTML chapter has corresponding IDs:
<p id="sentence1">Once upon a time, there was a fox.</p>
<p id="sentence2">The fox was very clever.</p>
The SMIL says "when the audio plays from 0 to 3.5 seconds, sentence1 should be highlighted."
Authoring Workflow
The synchronization data can be created several ways:
Manual: open the audio in Audacity, identify sentence boundaries by ear, write timestamps into SMIL by hand. Slow but precise.
Forced alignment: tools like Aeneas, Gentle, or DSAlign automatically align text and audio. Output SMIL or JSON. Very fast but accuracy varies with audio quality.
Author tools: dedicated ePub3 authoring tools (Sigil, Calibre with plugins, BookBaker) provide UI for syncing. Best for complex layouts.
For most production: forced alignment with manual review. Tools like Aeneas process an entire chapter in seconds, producing SMIL files that need cleanup but are 90% correct.
The media-overlay="chapter1-smil" attribute on the chapter item tells readers that synchronization data exists.
For each chapter, repeat this pattern. Some books synchronize the whole book to one audio file; some have per-chapter audio.
Audio Format Choices
Format
Compatibility
Notes
MP3
Universal
Default choice for ePub3
AAC (.m4a)
Most readers
Apple Books prefers this
Opus
Modern readers
Smaller files
FLAC
Some readers
Larger files
For Apple Books: AAC at 64-128 kbps mono produces clean audio. For Google Play Books: MP3 at 64-128 kbps. For mixed delivery: MP3 has the broadest support.
A passing EpubCheck doesn't guarantee good user experience but catches structural errors that platforms reject.
For Apple Books specifically, additional validation:
Use Apple Books' own validator (BookBaker has integration)
Test on iPad and iPhone
Verify highlighting is per-sentence, not per-paragraph for children's books
Reader Support Matrix
Reader
Media overlays
Accessibility integration
Apple Books (iOS, macOS)
Yes (full)
VoiceOver compatible
Google Play Books
Limited
TalkBack compatible
Adobe Digital Editions
Yes
Limited
Calibre Reader
Limited
Limited
Kobo Reader
Yes (newer firmware)
Limited
Amazon Kindle
No (KFX format different)
n/a
Thorium Reader
Yes
Strong accessibility
Vivlio (formerly Bookari)
Yes
Limited
Apple Books has the most polished media overlay implementation. Other readers vary widely in implementation quality.
For Kindle delivery, the workflow is different: KCC (Kindle Comic Creator) doesn't support media overlays. Audio Kindle books use a different mechanism through ACX.
Granularity Decisions
How small a text unit should be synchronized?
Granularity
Pros
Cons
Per-paragraph
Easy to author, less precise
Less educational value
Per-sentence
Standard for children's books
More authoring time
Per-word
Maximum precision
Very labor-intensive
Per-syllable
Karaoke-style
Specialty use only
For most read-along books: per-sentence. For language learning: per-word for the target language sections, per-sentence for context. For early readers: per-word with visual cues.
File Size Reality
For a 50-page picture book with audio:
Component
Size
Text (XHTML)
50-200 KB
Images (compressed JPG)
5-15 MB
Audio (MP3 64 kbps mono)
8-15 MB (for ~30 min)
SMIL files
50-200 KB
ePub package total
15-30 MB
For longer audiobooks integrated as ePub3: the audio dominates file size. Expect 50-150 MB for a full audiobook with synchronized text.
Pro Tip: Compress audio aggressively for read-along books. 64 kbps mono MP3 is sufficient for narration and saves bandwidth on mobile delivery. Higher bitrate doesn't improve perceived quality for spoken word.
Common Issues
Highlight doesn't follow audio: SMIL timestamps wrong. Re-run forced alignment, manually verify a few key transitions.
Apple Books shows "no audio available": SMIL or audio files missing from manifest. Verify with EpubCheck.
Tap-to-jump doesn't work: text element IDs in XHTML don't match SMIL references. Verify each <par>'s text src attribute exists in the chapter XHTML.
Audio plays but text doesn't highlight: highlighting CSS not injected. Most readers add highlighting CSS automatically; if missing, add epub:type="readaloud" to elements.
File rejected by Apple Books: media overlay structure doesn't match Apple's strict requirements. Use Apple Books' own validation tools.
For broader ebook format conversion, see our ebook converter.
Frequently Asked Questions
Can I add media overlays to an existing ePub3?
Yes. Add the SMIL files, audio files, update the manifest, add media-overlay attribute to chapter items. Re-zip the ePub. Validate.
Do media overlays work in Calibre?
Limited. Calibre's reader displays the audio and text but the synchronization quality varies. Test on actual delivery readers.
Can I have multiple voices/narrators?
Yes. Each <par> can reference a different audio source. For dialogue, alternate audio sources between speakers.
What about ASL or sign-language video for accessibility?
Media overlays support video too. The <audio> element can be <video>. ASL versions of children's books use this for deaf accessibility.
How long does it take to author a 30-page book?
With Aeneas forced alignment: 1-2 days for synchronization, more for careful review. Without automation: 1-2 weeks. The audio recording itself is separate (typically 1-3 hours per 30 pages).
Can I author ePub3 with media overlays in InDesign?
InDesign's ePub3 export doesn't include media overlays directly. Export the ePub from InDesign, then use Sigil or a programmatic tool to add SMIL synchronization.
For ePub3 with media overlays: use Aeneas or similar forced alignment for initial synchronization, validate with EpubCheck, target per-sentence granularity for children's books, deliver MP3 64 kbps mono audio. Apple Books has the strongest reader support. Our ebook converter handles ePub-to-other-format conversion if your delivery target needs PDF or KFX.
