A playlist generator for mpd, using metadata from EchoNest.
This assumes that music is stored in mp3s, with ID3 metadata, grouped in directories by album, with albums grouped in directories by artist.
The number of calls made to EchoNest is huge when first run (the scan and link scripts; even though the calls are aggressively cached it takes days to get the data for my music collection). So this is not a good use of EchoNest and won't scale well.
Because of the above (and, you know, laziness), this project is not packaged for easy use. On the other hand, I think it's pretty awesome and use it all the time.
So if you want to use this code, and are happy getting a clone running (it's pretty much trivial if you're an experienced Python programmer) please email me at andrew@acooke.org for the EchoNest ID (I think the idea is that the program should use a single ID in the API; it is designed to self-throttle if multiple users are running).
-
scan.sh- This reads the directories below the root (set in the configuration file), checking for new data. ID3 data for new tracks (and, randomly, some old tracks) are read and cross-referenced against the EchoNest database. -
link.sh- This checks for links between the artists found, using EchoNest's related artists. If new artists were found via scan.sh then all artists are updated, otherwise only a random selection. -
update.sh- Calls the above two scripts; expected to be called each night to update the local database as necessary. -
reset.sh- Wipes the database (except for the EchoNest cache).
A note on caching - all calls to EchoNest are cached and expire over a random period (between 0 and 30 days). This, together with random re-processing, means that updated data in EchoNest is slowly absorbed into the local database (typically within a month).
play.sh watches the mpd queue and takes action in two cases:
-
when the queue is empty, a new, random track is added (but not played);
-
when the queue contains a single entry a new, related track is appended.
The intention is that this script runs continually in the background. Selecting a random "starting point" is then possible by clearing the queue. Pressing play then generates a continuous playlist, song by related song. But if other music is selected then the functionality is automatically suppressed (until the final track is played, at which point another will be added).
Popular tags from EchoNest are also stored in the database. These can be used as follows:
-
ushow- Displays the tags (and other info) for a given artist. -
uwho- Shows the artists that match a given tag. -
uadd- Adds random tracks from artists that match the given tag.
Tags can be combined and excluded with a - prefix. So uadd 20 blues funk -pop will add 20 tracks from artists that are tagged with "blues" and "funk",
but not tagged with "pop".
unext - Will list files, given various constraints. So this can be used
to generate a long playlist, or to generate the "next" song in a related list.
transfer.sh will create a directory of artists and albums with links to
tracks, selected at random, so that the total is of a given size, that can
then be copied to a PMP (iPod etc).
This software includes a copy of Stagger which appears to be no longer maintained, and does not install. The copyright remains with the initial author, Karoly Lorentey. See LICENCE file in the stagger directory.