[Playlist] Mission questions

Robert Kaye rob at metabrainz.org
Sun Nov 29 12:56:33 UTC 2020


On Sat, Nov 28, 2020 at 10:30 am, Lucas Gonze <lucas at gonze.com> wrote:
> After long consideration, I believe the question that matters is the 
> existence of non-siloed music. Should it exist? Why?

Yes, they should.

As you said, techlash is one aspect and the other is that music tools 
are finally being worked on again. MusicBrainz Picard is seeing an 
uptick of activity, we recently discovered and are embracing Funkwhale 
(federated, self hosted music collections with streaming) and I've 
started paying attention again to my own music collection. I'm growing 
disenchanted with Spotify -- tracks appear, I fall in love with them 
and then they disappear without notice. Large chunks of what I love are 
not there and the utter and complete lack of interoperability, is the 
ultimate lock-in. I'm starting to get fed up.

With the ListenBrainz project we've been working flat out on building 
the a recommendation engine toolkit. Track similarity, collaborative 
filtering, super easy to use toolkits with the heavy lifting already 
done -- we're hoping to kick-start open source music recommendations. 
We're starting to produce results that are quite listenable, after only 
5 years of working hard. ;)

And to deliver generated playlists, we needed to add playlist 
capabilities to ListenBrainz. Last weekend we had an online hack-day 
where we had 5 people work through the weekend to add collaborative 
playlist support to LB. We'll deploy this before year end and then 
we'll start generating playlists (similar to discovery weekly and daily 
mixes) for users who scobble their listens to ListenBrainz (if you 
scrobble to us or import your last.fm history, the next monday we'll 
have some recommendations for you) We're working hard to become an open 
source mix of the Echo Nest and Last.fm and we're making great strides 
finally.

We realized we need a transport layer playlist format and we quickly 
agreed to use JSPF for this! We're using it as an exchange format and 
as our internal format as well --
we've defined two extensions for JSPF for our internal fields. I'll 
post a request for feedback in the coming days, to make sure we're on 
the right track.

> Given techlash, what is the point of interoperability?

We've also been working on our mapping game and making it super easy to 
take an artist name and a recording name and to find a suitable release 
in MusicBrainz. Have a look at our latest iteration on what will soon 
be the MusicBrainz ID mapping tool:

  <https://datasets.listenbrainz.org/acrm-search?query=sun+shines+tv>

For some tracks you don't need a lot of metadata to get the right 
release, for others you need to give more. The text -> MBID mapping is 
our first effort and MBID -> Spotify mapping will be the next. These 
mappings will prove to be important for getting out recommendation work 
exposed in tools like Funkwhale, which is sorely lacking 
recommendations and intelligent playlists.

I think sometime in 2021 we'll see a point where open source tools for 
federated music networks will reach some level of parity with the likes 
of Spotify.

Unsurprisingly, we hope that MBIDs will be the point of 
interoperability.

--

--ruaok

Robert Kaye     --     rob at metabrainz.org <mailto:rob at metabrainz.org>   
  --    http://metabrainz.org <http://metabrainz.org/>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xiph.org/pipermail/playlist/attachments/20201129/8f2db9bf/attachment.html>


More information about the Playlist mailing list