Why Can’t Music Streaming Services Give Good Recommendations for Classical Music?

I regularly use Apple Music, and sometimes the recommendations I receive in the For You section are spot on. They learn from what I listen and what I love (though I don’t love tracks or albums very much), and they recommend music by the same or similar artists, or from similar genres. On any given day, I’d say a quarter of their picks are things that I really would like to listen to. And I think that a 250 batting average for this type of recommendation, which is all done by algorithm, is pretty good.

However, when they recommend classical music, they tend to strike out a lot more. Last night, I listened to an album of Schubert’s piano trios, and this morning, I see these recommendations:

Classical recommendations

It’s fair to say that I’d be potentially interested in listening to many if not most of these recommendations, but are they really “like Schubert: Piano Trios, Op. 99 & 100?” No, not really. There are two recordings of violin concertos, an opera, some vocal music (Monteverdi’s Vespers), and some solo piano music.

What would be “like” those Schubert piano trios? Perhaps other chamber works, such as piano trios by Haydn or Beethoven. Maybe some string quartets by Schubert, Beethoven, or other Romantic composers. Or some other music by Schubert: his piano music, lieder, etc.

It’s not clear why these recommendations were chosen. With pop, rock, or jazz, the recommendations tend to be based on the artists performing the music, whereas here, this isn’t the case. None of the three artists who performed the Schubert trios I listened to (Andreas Staier, Daniel Sepec, and Roel Dieltiens) are present in the recommendations. Two of the recommendations are on the same label, Harmonia Mundi, and, in classical music, that can a good reason to recommend music, as independent labels do have a specific character. But I scratch my head to try to figure out how these recommendations were chosen.

The Igor Levit set is in my iCloud Music Library, and I have listened to it before, but I don’t know any of the other recordings. The only commonality I find is that the Schubert I listened to was released in 2016, and five of the seven recommendations were released the same year, with two others in 2014 and 2015.

It isn’t easy to tailor recommendations for classical music, and I suspect that Apple Music is simply looking at what other people who have the Schubert recording in their libraries are listening to, or what’s in their libraries, similar to the way the Genius feature works. Providing better classical recommendations would require additional metadata for classical recordings, beyond just the “classical” genre. There would need to be metadata for eras (Baroque, Romantic, etc.), ensemble sizes (trios, quartets, orchestras, etc.), and styles.

The classical market is too small for the big streaming services to provide this sort of recommendation, and other players, such as Idagio and Primephonic, are entering the field in an attempt to do so. This is probably not something that can be done by algorithm, in part because of the absence of extended metadata specific to classical music.

To be fair, a bit of browsing on Apple Music allows me to find plenty of classical music, but I really would like the kind of recommendation that pushes me in the right direction, especially for composers that I don’t know well. I’m not that interested in paying for another streaming service, because that sort of fragmentation with music is just an annoyance. But I wish the big streaming services – Apple Music, Spotify, and Amazon – would take classical music seriously.

Amazon Launches Amazon Music HD (and Lies a Lot)

Amazon today announced Amazon Music HD, an extension of their paid music streaming service offering 50+ million tracks in lossless FLAC format, and “millions” of tracks in high-resolution formats. (It’s also available from Amazon UK.)

Amazon, the music streaming service that is probably not used by that many serious music fans, hopes that this can get them to be a serious player in this market. But I’d expect that many if not most people who use Amazon to stream music probably use an Alexa device, in most cases devices where lossless audio won’t sound better than the MP3s that they serve.

Amazon has used all the audiophile tropes to try to sell their service, and this graphic sums them up:

Amazon music hd

First comes the stair-step graphics attempting to suggest that higher sample rates are better. Amazon has shamelessly ripped this off from Qobuz, who has been using this graphic for more than two years.

Qobuz lies

They then discuss bit rates. MP3 is “up to 320 kbps,” or what most people can hear correctly. But for the “high definition” audio – lossless, or CD quality – they say “up to 850 kpbs.” Anyone who understands lossless compression knows that the bit rate of a lossless file depends on the density and volume of the music, and higher bit rates are not better. In fact, it’s not uncommon for lossless files to have bit rates above 1,000 kbps, such as with this Clash album:

Clash bit rates

Or even well below 320 kbps, as with this album of piano works by John Cage:

Cage bit rate

And for “ultra HD,” or what is more commonly known as high resolution, saying “up to 3750 kbps — more than 10X the bitrate of standard streaming services” is disingenuous at best. If you have magical bats’ ears, you might hear the difference, but whether the bit rate is 3750 kbps or half that makes little difference if you don’t have high-end audio equipment, and especially if the music isn’t mastered well.

“So,” you are thinking, “Isn’t it time for Apple to offer something similar?” I doubt it. While providing lossless streams would fit well with the niche the company is trying to create for the HomePod (a mono device, mind you, where lossless or high-res music won’t provide a full stereo experience), most people who listen to music don’t care, and it is not very useful for mobile devices. I’m sure that one day Apple will offer a lossless plan for Apple Music, but I can’t see it as a priority.

In any case, if Apple does offer lossless streaming one day, I hope they won’t use the same type of deceptive language that Amazon is using.

Record Labels Splitting Long Tracks into Multiple Tracks to Maximize Streaming Income

The music streaming payment model is optimized for popular music: short songs, three, four, five minutes long. Record labels are paid by song streamed, not by the amount of time the music plays. An hour of a three-minute song counts as 20 plays, whereas if it’s a four-minute song, it only gets paid for 15 plays.

In an attempt to hack this system, some record labels – notably for classical music – are splitting music into multiple tracks. You won’t see this on, say, your standard symphony, where, while it would be possible to split four movements into ten or more, but you will see it on other works, ranging from long vocal works to non-standard classical pieces.

Here’s on example: Max Richter’s eight-hour Sleep. If you buy this from the iTunes Store, you will get 31 tracks, ranging in length from 2:46 to more than 33 minutes. But if you stream it on Apple Music, here’s what you see:

Sleep

That’s right, it’s 204 tracks, most of which are under three minutes. By splitting the music this much, the record label – Deutsche Grammophon – gets more than six times as much money than if it were in the original 31 tracks.

Each of the original tracks is named, with a part number at the end of the name.

This is a cynical way to hack the music streaming payment process, but I do feel that this system unfairly handicaps classical and jazz labels, along with some jam rock and other forms of music – Indian classical, for example. Streaming income should be paid by duration rather than by song, or there should be multiple tiers according to the length of tracks. It’s a shame that record labels have to resort to this sort of system to get paid fairly.

The Next Track, Episode #90 – Liz Pelly on Streaming Muzak and Playlists

The Next Track Blue Flat Button2 400pxJournalist Liz Pelly talks about streaming muzak, Spotify, playlists, and the future of streaming.

Listen to The Next Track: Episode #90 – Liz Pelly on Streaming Muzak and Playlists.

Find out more, and subscribe to the podcast, at The Next Track website. You can follow The Next Track on Twitter at @NextTrackCast, to keep up to date with new episodes, and new articles from the website.

The Next Track, Episode #77 – How to Get the Most out of Apple Music

The Next Track Blue Flat Button2 400pxDoug and Kirk discuss Apple Music, how it’s changed the way they listen to music, and how to get the most out of the service.

Listen to The Next Track: Episode #77 – How to Get the Most out of Apple Music.

Find out more, and subscribe to the podcast, at The Next Track website. You can follow The Next Track on Twitter at @NextTrackCast, to keep up to date with new episodes, and new articles from the website.

The Next Track, Episode #50 – Paul Kafasis of Rogue Amoeba Software on Streaming Music in Your Home

The Next Track Blue Flat Button2 400pxDeveloper Paul Kafasis joins us to discuss in-house streaming: AirPlay, Bluetooth, and Google Cast.

Listen to The Next Track: Episode #50 — Streaming Music in Your Home.

Find out more, and subscribe to the podcast, at The Next Track website. You can follow The Next Track on Twitter at @NextTrackCast, to keep up to date with new episodes, and new articles from the website.

BBC Radio 3 Testing Lossless FLAC Streaming

BBC Radio 3, the classical station of the BBC, is testing streaming in lossless format using FLAC. If you’re in the UK, you can try it out now. You’ll need to use Firefox, though some hi-fi streaming servers will be updated to offer access to this streaming quality. Cambridge Audio is looking for beta testers for this service if you have one of their devices.

The BBC discusses this project here. It looks as though the trial will last for at least five months.

If you are in the UK, and listen to classical music, try it out. It’s hard to perform a proper A/B test, but you could open two browsers – Firefox or another – and play from one, stop, then play from the other, to see if you hear the difference. If you try this, drop a comment below about your experience.

Streaming Music Services, From Most Screwed to Least Screwed – Gizmodo

[…] looking at the landscape, streaming services in general are kind of fucked. The payment model for streaming music services makes it very difficult for companies to actually make money. In addition to infrastructure costs, payroll, and marketing, music licensing fees paid to the labels mean that the profit margin for subscription streaming is often negative. Profitability, even for the biggest players, is still largely a pipe dream. As a result, it’s a game where only services with a lot of funding and a lot of users (Spotify), or a lot of backing by a parent company and a lot of users (Apple Music), are relatively safe. At least for now.

But who is the most fucked? Who is just sort of fucked? Who has the best shot at surviving? Here is a ranking of streaming music companies, ranked from most to least fucked.

A good overview of the many streaming music services, including a bunch most people have never heard of.

Source: Streaming Music Services, From Most Screwed to Least Screwed

The Next Track, Episode #43 – Streaming Music 2.0

The Next Track Blue Flat Button2 400pxChris Connaker joins us to discuss how music streaming service can make the next step.

Listen to The Next Track: Episode #43 – Streaming Music 2.0.

Find out more, and subscribe to the podcast, at The Next Track website. You can follow The Next Track on Twitter at @NextTrackCast, to keep up to date with new episodes, and new articles from the website.

Sprint Acquires 33 Percent of TIDAL and Creates Game-Changing Partnership | Sprint Newsroom

Global music and entertainment platform TIDAL and Sprint (NYSE: S) announced today an unprecedented partnership that will soon give Sprint’s 45 million retail customers unlimited access to exclusive artist content not available anywhere else.

Uh, okay. Sure, they have some exclusives, but nobody cares that much.

This said, I’ve been predicting for a while that we’d see more telecom companies either buying streaming services, or working with them to bundle those services with phone contracts. One hurdle to successful streaming music is the cost of data, and presumably Sprint will either make Tidal data free, or charge much less for it.

However, Sprint, being almost entirely US-only (they have some presence in Latin America), won’t be able to offer Tidal streaming to users around the world. Perhaps they’ll make deals with other carriers, but this puts Tidal in a silo. Granted, if Sprint gets enough of its 45 million customers to opt in, that’ll be good for Sprint. Their goal is, obviously, to use Tidal – either free or at a discount – as a carrot toward locking customers into a phone contract. Because if you’re committed to a music streaming service, you’re less likely to switch to another carrier. (Or so they hope.)

Source: Sprint Acquires 33 Percent of TIDAL and Creates Game-Changing Partnership | Sprint Newsroom