What are we building?

At the end, Audiobookshelf 2.36.0 runs in Docker behind your Traefik at https://YOUR_DOMAIN with a valid Let’s Encrypt certificate. You get two libraries: one for audiobooks (with chapters, covers and metadata from Audible) and one for podcasts that downloads new episodes automatically. Listening progress is stored server-side in your account – you continue on your desktop exactly where your phone left off.

Audiobookshelf isn’t meant for music – Navidrome is the right tool for that, and Jellyfin for films and series. All three run happily side by side.

Audiobookshelf is pleasantly lightweight: the container uses around 55 MiB RAM when idle, the image is 469 MB. Disk space is needed almost exclusively by your audio files.

Prerequisites

• A server running Debian 13 with Docker (tested on a netcup VPS with Docker 29.6.1 and Compose v5.3.1)
• A running Traefik with its proxy network, as set up in Traefik as a reverse proxy
• A subdomain pointing at your server – see Connecting a domain to your server
• The basics of Docker Compose and bind mounts
• Your audiobook files (MP3, M4B, FLAC …). If you have none yet: LibriVox offers public-domain audiobooks in many languages.

Step by step

1. Create the directories

Audiobookshelf separates four data areas – and that separation is exactly what makes backups easy later: config (database, accounts), metadata (covers, cache, internal backups) and your media in their own folders.

mkdir -p /opt/audiobookshelf/{config,metadata,audiobooks,podcasts}

The image starts as root by default; with the user: "1000:1000" above, the process runs under UID/GID 1000 instead. The folders therefore have to belong to that user, otherwise Audiobookshelf can neither scan nor store podcasts:

chown -R 1000:1000 /opt/audiobookshelf

2. Add audiobooks – the folder structure decides

Audiobookshelf derives author and title from the folder structure and enriches them with the tags inside the files. Stick to Author/Title/ and the mapping is right from the start:

/opt/audiobookshelf/audiobooks/
└── Franz Kafka/
    └── Ein Landarzt/
        ├── landarzt_01.mp3
        ├── landarzt_02.mp3
        ├── landarzt_03.mp3
        └── landarzt_04.mp3

Several files in one title folder are combined into a single audiobook with chapters – not into four separate books. After copying, fix the ownership:

chown -R 1000:1000 /opt/audiobookshelf/audiobooks

3. Write the Compose file

Create /opt/audiobookshelf/compose.yaml and replace YOUR_DOMAIN with your real subdomain:

services:
  audiobookshelf:
    image: ghcr.io/advplyr/audiobookshelf:2.36.0
    restart: unless-stopped
    user: "1000:1000"
    environment:
      TZ: Europe/Berlin
    volumes:
      - ./config:/config
      - ./metadata:/metadata
      - ./audiobooks:/audiobooks
      - ./podcasts:/podcasts
    networks: [proxy]
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.abs.rule=Host(`YOUR_DOMAIN`)"
      - "traefik.http.routers.abs.entrypoints=websecure"
      - "traefik.http.routers.abs.tls.certresolver=le"
      - "traefik.http.services.abs.loadbalancer.server.port=80"

networks:
  proxy:
    external: true

Three details matter here:

• loadbalancer.server.port=80 – inside the container Audiobookshelf listens on port 80. The 13378 from the official example Compose is only the port you publish when running without a reverse proxy. We publish no port at all: Traefik reaches the container through the proxy network.
• The pinned version 2.36.0 instead of latest – updates then arrive when you plan them, not in the middle of the week.
• user: "1000:1000" matches the ownership from step 1.

Start the stack:

cd /opt/audiobookshelf && docker compose up -d

Check that the container is running:

docker compose ps

You should see Up, not Restarting:

NAME                              STATUS         IMAGE
audiobookshelf-audiobookshelf-1   Up 8 seconds   ghcr.io/advplyr/audiobookshelf:2.36.0

4. Verify TLS before you click

Traefik needs a few seconds for the certificate. The status endpoint tells you whether everything is in place – it answers without a login:

curl https://YOUR_DOMAIN/status

{"app":"audiobookshelf","serverVersion":"2.36.0","isInit":false,"language":"en-us",
 "authMethods":["local"],"ConfigPath":"/config","MetadataPath":"/metadata"}

Two things are now certain: 2.36.0 really is running, and isInit:false means the server is still waiting for its first account. If you get a certificate error here, look at the Traefik logs first – not at Audiobookshelf.

5. Create the first account

Open https://YOUR_DOMAIN in your browser. On the first visit Audiobookshelf asks for a username and password for the root account, i.e. the administrator.

Afterwards you’ll notice the address jumps to https://YOUR_DOMAIN/audiobookshelf/. That is intentional on the project’s side: the web interface is built with the fixed router path /audiobookshelf. There is a ROUTER_BASE_PATH variable, but it does not help at runtime – the path is baked into the pre-built frontend inside the image (tested: with ROUTER_BASE_PATH: "" the interface still loads from /audiobookshelf/_nuxt/…). Visiting https://YOUR_DOMAIN still works, your bookmarks just carry the suffix.

6. A note on the interface language

The interface follows a per-browser setting, not a per-account one: Audiobookshelf stores it locally and does not save it on the user. You pick it under your username in the top right corner (Language), and you have to set it again on every new device.

7. Create the audiobook library

Under Settings → Libraries you create the first library:

• Media Type: Books
• Library Name: e.g. Audiobooks
• Metadata Provider: pick the Audible catalogue for your language – Audible.com for English titles gives noticeably better matches than the default Google Books. iTunes, FantLab.ru and the other Audible country catalogues are available too.
• Folders: enter /audiobooks and confirm with Enter. That is the path inside the container – thanks to the bind mount it points at /opt/audiobookshelf/audiobooks.

Then click Create.

8. Trigger the first scan yourself

This is where most people stumble: the library stays empty at first, and the interface says so.

The first scan does not start on its own – not even after minutes (measured). In the library view, use the scan action offered next to that notice. With four files and 14 MB the scan finished in under four seconds, after which the audiobook appears with its cover:

From now on Audiobookshelf watches the folder itself: newly copied files are picked up, and a manual scan is only needed when creating the library.

9. Listening – chapters, speed, sleep timer

Clicking the cover opens the detail page. The four MP3s have become one audiobook with four chapters; duration and size are shown at the top:

The player at the bottom stays put while you navigate. It offers playback speed (1.0x), jump marks, bookmarks, the chapter list and a sleep timer. The progress – 0:06 / 28:52 in the screenshot – is stored server-side in your account, not in the browser. That is precisely why you self-host this thing.

10. Subscribe to podcasts

Podcasts need their own library, because Audiobookshelf treats them differently: episodes, feeds, automatic downloads. So go to Settings → Libraries → Add Library again, this time with:

• Media Type: Podcasts
• Name: e.g. Podcasts
• Metadata Provider: iTunes
• Folder: /podcasts

In the new library, click Add. The input field takes either a search term – then the server queries the iTunes catalogue – or the feed URL directly:

When subscribing you decide whether new episodes are downloaded automatically and how many Audiobookshelf keeps. Keep disk space in mind: a weekly podcast with 60-minute episodes costs roughly 30 MB per episode at 64 kbps and 55 MB at 128 kbps – so about 1.5 to 3 GB per year.

11. Add more users

Listening progress belongs to an account – so create separate users for your partner or kids instead of sharing the root account. That happens under Settings → Users → Add User.

There are three account types: root (only once, the first account), admin (may change settings and libraries) and user for the normal case. A freshly created user account gets these permissions:

download: true       – may download titles (for offline listening in the app)
update: false        – may not change metadata
delete: false        – may not delete anything
upload: false        – may not upload files
accessAllLibraries:  true    – sees ALL libraries
accessExplicitContent: false – titles flagged explicit stay hidden

The second-to-last line is the important one: by default a new user sees all libraries. If the kids should not browse the grown-up audiobooks, turn accessAllLibraries off in the user dialog and pick the allowed libraries individually.

12. Apps for your phone

Everyday use happens on the phone. There is an official Audiobookshelf app for Android (Play Store and F-Droid) and for iOS. On first launch you enter:

• Server address: https://YOUR_DOMAIN – with https://, without a path
• Username and password of your account

The app then downloads titles for offline listening and reports progress back to the server. On iOS there are also alternatives such as Plappa or ShelfPlayer that use the same API.

When things go wrong

Your problem is not listed? Search all error messages →

Maintenance & backups

Updates. Audiobookshelf releases often, roughly a new minor version per month. Because the version is pinned, you update deliberately: raise the tag in compose.yaml, then

cd /opt/audiobookshelf && docker compose pull && docker compose up -d

Before jumping across several versions, read the release notes – and take a fresh backup. How to do this systematically is described in Keeping your Docker stack up to date.

The built-in backup – switch it on first. Audiobookshelf can write archives itself, but it does not do so out of the box: under Settings → Backups the automatic backup toggle is off by default.

Switch it on and check the two limits below it, because both are set tight:

• Number of backups to keep: 2 – older ones are deleted. For „I only notice a problem two weeks later" that is too few; 7 to 14 is more realistic.
• Maximum backup size: 1 GB – if the archive grows larger, the backup aborts. With many covers in /metadata that can happen; 0 means „no limit".

The archives end up in /metadata/backups, in our case /opt/audiobookshelf/metadata/backups. They contain users, listening progress, library item details, server settings and the images from /metadata/items and /metadata/authors. The interface states it clearly itself: files from your media folders are not included.

Restoring works on the same page via the upload option – worth knowing, because it reduces a move to a new server to two steps: upload the archive, copy the media.

Real backups. That gives you a clear split:

• config/ and metadata/ are small (480 KB and 112 KB here) and belong in a daily backup – that is where the actual work lives: accounts, progress, metadata.
• audiobooks/ and podcasts/ are large and rarely change. They need a backup too, but on a different schedule – podcast episodes can even be classed as „re-downloadable any time".

Restic handles both reliably. Once a quarter, verify with a real restore that the backup actually comes back.

curl -X POST https://YOUR_DOMAIN/api/libraries/YOUR_LIBRARY_ID/scan \
  -H "Authorization: Bearer YOUR_TOKEN"

Keep an eye on disk space. Audio files and automatically downloaded episodes are the only real space eaters:

du -sh /opt/audiobookshelf/*

14M     /opt/audiobookshelf/audiobooks
480K    /opt/audiobookshelf/config
112K    /opt/audiobookshelf/metadata
4.0K    /opt/audiobookshelf/podcasts

If it grows beyond your disk, additional block storage for the media folders is the cheaper answer than a bigger server – the rest of Audiobookshelf stays tiny.