Configure your media server, authentication and advanced settings.
Connect one or more media servers (Jellyfin, Emby, Navidrome, Lyrion, Plex, even several of the same type). Analysis covers every configured server: the default server runs first, then each remaining server in turn. Songs shared by several servers are recognized by their audio fingerprint and analyzed only once, and songs that exist on just one server are analyzed too, so the catalogue is the union of all your libraries. Alignment sweeps map the shared content across servers as analysis progresses. Use the server switcher at the bottom of the menu to pick which server the interface works against: playlists and searches then target that server, and tracks it does not have are skipped.
No servers configured yet.
An admin account already exists. To add, remove, or change admin and normal user accounts, go to the Users page.
Choose the models that run during analysis. Turning a model off also turns off every feature that depends on it: open "What it does" under a model to see which pages use it and why. The small bar under a model shows how much of your library that feature can find right now; it fills up as the analysis runs.
We suggest keeping it on even if you never open Text Search: the mood scores feed the other pages, and new songs get none without it.
A fallback: if a song has no lyrics on the music server and none from the configured Lyrics API, Whisper transcribes them from the audio. The song can then still be found in Lyrics Search.