Convert MP3 to M4B for Audiobookshelf

Audiobookshelf reads a chaptered .m4b the way you'd hope: one file, chapters from the embedded markers, cover and metadata from the tags. Kataribe builds that file from a folder of MP3s in your browser, before anything touches the server.

Moving a collection into Audiobookshelf is when the folders show their age. A book of 60 MP3s works, but it's 60 files to copy, chapter names that are whatever the tags said, and a cover that may or may not be picked up. Drop the 60 files here, check the pre-filled title, author and narrator, tidy the chapter names, and put one .m4b in the library folder. On scan, Audiobookshelf reads the chapters, cover and metadata straight from the file, with nothing to match by hand.

Because the encode happens on your laptop rather than the server, it doesn't matter if Audiobookshelf runs on a Raspberry Pi or a low-power NAS. One book per build; for a whole shelf, work through them one at a time, and for very large sets (10 GB or more of input) build in parts.

How to convert MP3 to M4B for Audiobookshelf

  1. Drop the book's MP3s and confirm the order. Track tags order the files when every file has one; otherwise natural filename sort applies. Audiobookshelf trusts whatever order the chapter list has, so verify it here, especially for rips where one disc was tagged differently from the others.
  2. Name chapters the way the chapter panel should read. Audiobookshelf shows embedded chapter titles verbatim and uses them for the sleep-timer and skip controls. Titles come from ID3 titles or cleaned filenames; a set titled "Track 01" through "Track 60" works but says nothing, so at least name the chapters that matter.
  3. Fill the fields, and know which one is missing. Title, author, narrator and year are pre-filled from the MP3 tags when present; check them, since a ripper's album artist is often the narrator and not the author. Series is not a field in Kataribe and is added in Audiobookshelf after the scan.
  4. Bitrate for a server that streams. The player streams the file at its stored bitrate, so 64 kbps means 64 kbps over the network and about 29 MB per hour on the disk. On a NAS holding hundreds of books that adds up; 64 for narration, 96 for productions with music, nothing higher.
  5. Cover, build, place. Embed the cover here (JPEG or PNG). On scan, Audiobookshelf uses an embedded cover when the book folder has no image file of its own. Download the M4B and drop it into Author/Title/ in the library; the placement and scan steps are below.

How Audiobookshelf reads chapters and metadata

On scan the server probes each audio file with ffprobe. From an M4B it gets the chapter list with start and end times and titles, the duration, the codec and the embedded tags. Chapters are stored on the library item and shown in the player; they can be edited later in the chapter editor, and that edit lives in the database, not the file.

Tags map to fields: title, artist as author, composer as narrator, date as publish year, genre as genre. The library's metadata precedence decides whether those tags override the folder names, an OPF file or an online match, so a library set to prefer folder structure will show Author/Title from the path and let the tags fill in the rest.

One file per book on a small server

A single file per book simplifies everything downstream of the scan. Backups are one file per title, the mobile apps download one file for offline listening instead of sixty, and moving a book between libraries is a single move. Chapter data lives in the file, so a rebuilt database or a fresh server recovers it on the next scan.

Audiobookshelf can also merge files into an M4B itself, and that is a fine route on a capable server. Doing it here first means the chapter titles and tags are right before the file ever lands in the library, and the server's CPU stays free for playback and scanning.

Naming the folder versus tagging the file

Audiobookshelf's folder convention (Author/Title, optionally Author/Series/Sequence - Title) carries information the file can't: series and sequence in particular. The tags Kataribe writes carry what the folder can't: narrator, year and the chapter titles. Use both; they don't conflict, and the precedence setting resolves the overlap on title and author.

The output file is named Author - Title.m4b by default. Audiobookshelf doesn't need the filename for metadata when the folder is structured, so rename it or leave it; what matters is that it sits alone in its folder without the original MP3s beside it.

After you download the .m4b

Audiobookshelf

  1. Copy the .m4b into your library folder as Author/Title/Book.m4b (or upload it from the web UI).
  2. Run a library scan; Audiobookshelf reads the embedded title, author, narrator, cover, and chapter markers.
  3. Chapters appear in the player on web, iOS, and Android — no chapter files or metadata edits needed.

Frequently asked questions

Audiobookshelf can make an M4B itself. Why do it here?
It can, using the server's CPU, and on small hardware a long book is slow and ties the server up. Kataribe does it in your browser before upload, with chapter titles and metadata fixed first, so the file lands in the library finished.
Will Audiobookshelf pick up the chapters automatically?
Yes. Chapter markers embedded in the M4B are read on scan and appear in the chapter panel. The embedded cover is used, and title, author and narrator tags are read subject to the metadata priority set for the library.
Should I keep the MP3s as well?
Your choice; the M4B is a re-encode, so keeping the originals costs only disk. Don't put both in the same book folder, or the scan treats them as one book with duplicate tracks.
Does the file name matter?
Audiobookshelf reads author and title from the folder path first (Author/Title/book.m4b), so name the folders the way you already do. The .m4b itself can be called anything.
Does Audiobookshelf need M4B or can it use MP3s?
It plays a folder of MP3s fine, treating each file as a track and, when there are no embedded chapters, using the files as chapters. The M4B gives you one file to copy, back up and download to the mobile app, chapter titles you chose, and metadata read from the file rather than guessed from track tags.
Can I set the series in Kataribe?
No. There is no series field; the builder writes title, author, narrator, year and genre. Audiobookshelf takes series from the folder layout (Author/Series/Title) or from the item editor after the scan, so set it there.
Will Audiobookshelf transcode the M4B when streaming?
Normally no. AAC in an MP4 container is direct-played by the web player and both mobile apps, so the server only serves bytes. Transcoding happens when a client can't direct-play the source, which a Kataribe M4B doesn't trigger.
Are my audio files uploaded anywhere?
No. Kataribe runs FFmpeg compiled to WebAssembly inside your browser, so the files never leave your device. The engine (about 30 MB) downloads once and is cached; after that there is no account, no queue and no watermark.
How long does a full book take?
Encoding runs many times faster than real time on a modern laptop, but a full-length book still takes minutes, not seconds. Keep the tab open until the download appears. If every input is already AAC, the audio is copied rather than re-encoded, which is much quicker.
Is this really free?
Yes. Every input format, every bitrate, cover art and metadata, no sign-up, and nothing added to the audio.
Does it work with DRM-protected audiobooks?
No. DRM-protected files (Audible .aax and similar) won't open, and we don't help with that. Kataribe is for audio you already have in plain MP3, M4A, WAV, FLAC or OGG.