My guest today is Thomas Dose. Thomas is the head of Music Services at DR, the Danish Broadcasting Corporation. In this role, he works with a massive archive of physical and digital music files. In this episode, we talk about how DR organizes it all and what we can learn to help us manage our own music collections better.
Listen to the full conversation
- Thomas Dose on Twitter
- Jorge’s blog post about Idagio
- Apple Music
- The Dark Side of the Moon by Pink Floyd
- Acousting fingerprint
Read the full transcript
Jorge: Thomas, welcome to the show.
Thomas: Thank you Jorge.
Jorge: It’s great having you here. For folks who are tuning in and who don’t know about you, how do you describe what you do?
Thomas: Yeah, well, I’m Head of Music Services in DR. That is the national radio and TV Broadcasting Corporation. The short story of DR is that it is the equivalent of the BBC in UK. So we are a public service organization founded in 1925 and we basically provide news and culture, music, entertainment and such through TV and radio and online services and as well as having seven orchestras and a concert house.
So I work within this facility in what used to be the physical music archive and is now basically… We’re still managing the physical archive, but we’re also obviously we’re more data-driven now. So we are developing and operating a number of internal digital services and I work mostly as a product owner for these services and I spend basically all my time thinking about how to organize music and how to facilitate the music production of DR.
Jorge: That’s fantastic. I’m curious about your background in order to do that. Do you have training in music or…
Thomas: No. Actually and I don’t have an academic background. I was through the 90s and for as long as I can remember in the early part of my life, I wanted to be a musician or rather a producer. I grew up with open music genres, hip-hop and soul and R&B and a number of electronic music genres. And that was basically where my heart and head was for the first 25 years of my life. But eventually, it became clear that I was more focused on how music worked, I mean technically, than actually… And my talent didn’t really reach for a career in the arts. So during the 90s I transitioned into being a sound technician, at first as a trainee on Film School in Copenhagen and then working as a sound technician on a production company. And then parallel with that, I was working in an independent record shop and then from the mid-90s I was exposed to the web and became obsessed with all disciplines of that, including programming and the design. I was part of a quite lively design community in the late 90s in Copenhagen around music. We did a lot of interesting projects with driven radio stations and playlist services and so on and that basically led to the job. I’m holding now In the in the DR.
Jorge: You mentioned something that I took note of and I’m intrigued by, which is that you have digital archives and you also have physical archives.
Jorge: What kind of things do you store in the physical archives?
Thomas: Yeah, we don’t store as much anymore though. The department I’m working in has been systematically collecting music since 1949, and the physical archives that they consist of roughly about 900,000 physical units, that is records, which are shellacs, vinyl, CDs, and so on. But obviously for the last decade or so, we haven’t really added much to the physical archive. Only on those instances where a release is purely on physical, we will acquire that such. What else it’s all digital now. But we’re still very happy with the physical archive. It’s not collecting dust because the editorial units in DR are basically ordering digitization of older materials every day, and we handle those. And we digitize those from from vinyl and from shellac. And you would be surprised of the volume of music that is still not available on the mainstream streaming services. You think that it’s interesting that every piece of music recorded ever is on Spotify. It’s not nearly the case. So we’re still recording from from our physical archives.
Jorge: One of the differences between a physical archive on a digital archive is that by the very nature of the thing you’re storing, the physical disks or objects or whatever it is can only be on one shelf in one order at any given time. I’m curious, 900,000 sounds like a lot. Do you have a particular organization scheme for those?
Thomas: The only sensible way to organize collecting like that is to basically have it placed in the order they were acquired, and then have an identification to, and then catalog the all the items in order to find them. We do organize them by format. So we have twelve-inch vinyls on one part and then we have seven-inch vinyls and so on. But that is that’s the way. So basically when you go through the physical archive, it’s sorted in the order they came in which also makes it interesting because you can dive into periods of music that way. 70s, 60s… But that that’s how it was done. We acquired or developed our first electronic system in 1978. And from there on it’s mostly a true digital system we’ve been finding and organizing our music.
Jorge: That makes sense. And that actually is the primary reason that you and I are talking. I came across this company a few… I guess a couple of months ago, called IDAGIO. And I wrote a little blog post about it because it it intrigued me that it’s a service that is using the organization of information — a very particular type of information in this case, classical music — as a competitive differentiator against competitors such as Spotify or Apple Music. And I was very intrigued by this and a common friend of ours suggested that we talked about it. So I’m guessing that you have thoughts on the organization of digital music and in particular this challenge that companies like IDAGIO are looking to solve.
Thomas: Absolutely. Well, no doubt. And it’s long overdue. And I’d say that there are streaming reading alternatives to Spotify doing it the right way in terms of describing music not only as a track or a recording, but also the relationship between a recording and the original composition or work and which is basically the crucial component to IDAGIO’s business model. We’ve been doing that as well and DR for a few decades. We were somehow through coincidence lucky enough to not be organized within the program archives in DR and that is quite atypical in all other public broadcasting organizations I know of. In DR, the music archive is a department of the programming archives and the programming archives traditionally lean towards structuring their items the same way as the libraries would do. And through not having been in the program archives, organizations with basically been developing music data models from the 90s that supported this structure of having the composition and the recording separated, which is yeah, the main point is either of IDAGIO. You see a number of other streaming services going the same route as well. But Spotify, the large mainstream streams are still lacking in that department.
Jorge: What I’m hearing you say is that in classical music, perhaps more so than in other types of music, the distinction between the artist who composed the piece that you’re interested in, can be as or not more important than the artist who performed the piece.
Jorge: And in things like pop music, usually those two things are merged somehow. And if you’re thinking of listening to a song by someone like Taylor Swift, you just search for Taylor Swift. Whereas if you’re wanting to listen to music by Gustav Holst, there might be many many recordings by different orchestras and different conductors. And how does one deal with that distinction?
Thomas: Well in our case, our data model basically supports two types of composition. And one is, you could say, the normal type of composition where you have a title for the composition and then you would have composers and lyricists related to that. And the other type of composition would support sub-compositions, which is basically in one of the obvious example is you have a symphony which would have four movements and then and so those are the sub-compositions. And we are then able to relate each of these sub-compositions or movements to all the different recordings of this movement and this work. So you would have obviously the the classical editorial units in DR would need to search for the composer and then from there on see what recordings there were of this composition, including the orchestras, the conductors, choirs or soloists’ performance. And all this is not that different to the structure of of a composition but it’s just a crucial step to having these two entities that composition and recording available to you.
And also IDAGIO is basically exploring this model but really a lot of other genres could do. Well, it’s I mean, especially yes, you would want to explore jazz standards through the different performance recording one composition and also in pop music you would have remixes, you have live versions of the same composition. So really it benefits all genres of music but classical it’s just the obvious end to start having these solutions.
Jorge: Yeah, what I’m hearing you describe there is that it’s desirable to have some way of connecting a piece of music with perhaps variations on that piece or alternate takes or something like that.
Thomas: Yeah, that’s it. See I think you’re gonna… I think Spotify, as we agree on, is lacking, but now you actually can view the credits or at least the composers and the producers of most of the tracks at Spotify. They’ve just not made them searchable or they haven’t implemented this information in a very convenient way yet. So I think you’re going to see some of those developments in some of the mainstream services as well.
Jorge: I have a lot of thoughts hearing you describe this. I’ll just pick one because there’s so many avenues we could go down here. One is that there is certainly a distinction between someone who does this, someone like yourself who does this for a large centralized music archive, such as DRs, right? Like there’s this… That would strike me as a kind of one-off instance of a music catalog where you’re maintaining this catalog and you have to provide it with all these connections and all these organizations schemas so that the people who need the music can easily find it. And the other is what you’re describing with regards to services like Spotify, which is we as consumers of music or listeners of music who just want to listen to something.
Jorge: And we we don’t have necessarily very good ways of doing this. So I just wanted to draw that distinction between you know, the work that you’re doing as a kind of one-off for this very large…
Thomas: Yeah. I think the use cases are connected. It’s just that these types of solutions that were talking about has been basically only for professional use and we’re starting to see with examples like IDAGIO the solutions being fully being put into mainstream solutions as well. I mean we used to have when music was distributed physically, you would have credits obviously on cover sleeves and for a lot of people those were crucial information. And so I think the case of not having those information available as search parameters in streaming services, it’s probably just a case of a decade or two transitioning in from physical to streaming. And I think we’re going to see a lot of… I think I’m really excited about the next few years because we are seeing the signs now that both from the major label end of the business but also in the open source community, you see solutions providing a much more detailed mixed tastes on music, and I think IDAGIO is only the beginning. You have in streaming services now, obviously Spotify is still is the biggest one, surpassing 200 million users, but it’s not… You’re starting to see what you could call boutique solutions popping up catering for special needs. Most of this development, I think, has been driven by basically, the audio quality, the technical quality of the streaming. You’ve seen streaming services like Tidal and Qobuz basically carving out niches within the Hi-Fi communities and their price model is very much based on the technical quality of the streaming of the audio.
But you’re starting to see that these same services are also competing on the quality of search and metadata basically, which is one thing. The other thing is having these metadata available at first. And this is obviously been the biggest obstacle towards delivering these services. You had a music industry that has been for almost always not really been that tech savvy to to drive those kind of solutions. So it’s no coincidence that it’s tech companies coming in and and disrupting this this business. You also had a case where the music industry didn’t really care because they were making enough money, basically through selling the music. And what I’ve seen in the last 10 years or so with the crisis in and the devaluation, commercial devaluation anyway, of music, is that even the major labels are now starting to push for better metadata because that is one of the sources of revenue now for them. It’s metadata basically that provides them with the royalties from video or any other public use. So you’re starting to see the development from both the major industry and as well as the open source communities.
Jorge: When you say open source, are you referring to a piece of software that you install on your computer?
Thomas: No, I’m more thinking about some of the larger open source music databases. One important one is the musicbrainz.org Initiative, which is basically a catalog of music. And one is special with I think with the MusicBrainz is that data model is very sound and it’s quite advanced and it supports both the solutions we’ve been discussing as well as a lot of others. And MusicBrainz is is driving a lot of development in some of these areas mostly through products ingesting these data and then providing music streaming services on top. I was discussing the Hi-Fi communities one, very important product in that end is Rune which is basically a home streaming solution that supports both your local collection, sound files collection, and as well as your subscription to Tidal or streaming services like Qobuz. So you could basically combine your private collection with these streaming services and Rune lets them provide say a quite convenient interface based on basically MusicBrainz music metadata. So you would have all these search facilities that you’re like you’re talking about IDAGIO has as well.
Jorge: It sounds like it’s a service that solves a problem that I face . I’m old enough where I went through the transition from having music in physical media to having a digital collection and that meant digitizing a bunch of CDs, for the most part for me.
Jorge: And one of the challenges that I faced there is that… You were touching on this earlier with the notion of the sub compositions. So I have a lot of like electronic music and rock and stuff like that, more kind of pop music. And those are fairly straightforward because those are musical works that were produced for the most part during the era of recorded music, right? So for the most part of the composition fits into an album, this concept of an album. And I’m thinking of something likePink Floyd’s Dark Side of the Moon which I think for a long time was the biggest selling album and that’s a composition that is as long as an album because of the media that it was recorded for, right?
Jorge: And it made the transition to CDs because CDs were designed to accommodate most of these things. And when digitizing that, all that software kind of accommodated that very nicely.
Jorge: But then I also have a bunch of classical music CDs, and those include musical forms that are not easily fit into an album length thing, right? You have things like like piano sonatas which are much shorter than what fits into a disk. And therefore you have multiple of those in one plastic plastic disc. And the challenge for me when digitizing all the stuff was, do I like… Let’s say that I’m digitizing a disk that has a bunch of Beethoven piano sonatas. Do I save each sonata as its own work or do I save them as a set? Because I acquired it as a set, right? It’s been well over a decade since I did this, and I still haven’t resolved this. How does one solve this?
Thomas: I think what the problem is with these cases. There are no one perfect solution because each release will have its own way of organizing. You see so many variations of releases of the same work, and you have to support those. So I would for I think as a private music user I would definitely look into some of these solutions like Rune or others that have made the attempt to solve this for you. Because it’s always going to be a compromise with these types of metadata. I mean, that’s even with some of these factual information that everyone should be agreeing off. But but you see so many instances In classical releases that you would have several movements on the same track on the CD, you would have… So if you’re organizing based on your media, your CDs, you’re going to be in pain for a lot of time. I think the best way to go about it is basically to decide on one of the sensible solutions out there and then live with some of the problems. I personally, with my own with my own consumption I gave up on organizing my own music collection a few years ago once I saw that there were actually some very sensible solutions beginning to emerge. So I would I would I would pass that on to you as a key advice.
Jorge: Yeah, thank you for that. I actually did the same I so my trajectory went like this. I digitized my music then I signed up for this –I think it was called Apple music match — service, which basically gave me cloud based versions of the things that I already had.
Jorge: And one of the things that happened when I made that transition is that. For some reason Apple restructured some of the things that I had already digitized, especially around things like classical music that have these challenges we were talking about. And it it basically broke a lot of my careful organization. And I got so frustrated that I said, you know, I’m not going to bother with this.
Thomas: Yeah, because I remember that. I didn’t I was lucky enough to avoid that myself. But I remember that being the source of quite some scandal that a lot of music lovers basically ruined their collection and all their work of organizing the music for years through that solution. What I didn’t mention was one of the key benefits of using some of the Contemporary Solutions is that you basically only provide your sound files and through audio fingerprinting these services will then recognize and identify all your music and will envision your music files with all the correct metadata as well as album covers and cover sleeves if they’re available and so on. So it’s really that easy now to combine your carefully collected sound files through the years and then your student preferences. So I think it’s looking up.
Jorge: We’re nearing the end of our time together, but I can’t pass up the opportunity to ask you about audio fingerprinting. I mean, you said that and it just intrigues me. What is audio fingerprinting?
Thomas: Audio fingerprinting is the most popular popular technical way to identify an audio file.
It’s been popularized by Shazam. I don’t know if you ever use that service, but Shazam is a mobile app and it’s very popular, especially with younger demographic. It basically makes it possible if you have the app installed to open your microphone and then listen to a piece of music in a room or a restaurant or a club and then Shazam will recognize the music and tell you what it is.
The problem with Shazam is you’re only receiving the information about the artist and the title. But other services, like the MusicBrainz initiative I was just talking about, on top of MusicBrainz there is an audio fingerprinting service called AcoustID, which is the solution that most of these new products are using, including Rune.
So basically Rune through AcoustID will scan all your audio files analyze them and then match them in their database and then through that have access to the complete music database of MusicBrainz. So it does and it does not it’s not like all your fingerprinting is affecting the sound files itself.
We’ve been doing audio fingerprinting ourselves in DR and it’s basically been a huge game changer in the way we’re administering our music because it’s… Yeah, basically before we needed to have a very complex flow of metadata to complete production in order
Jorge: That’s fascinating and I’m sure that we could keep talking about this or so many things to to discuss in this field, but that feels to me like a good place to wrap up our conversation for today. Thank you for that. Just one final thing, if folks want to follow up with you, see what you’re up to, what is the best way for them to do that?
Thomas: Actually, I’m not that active on social media. I kind of was disillusioned by social media a couple of years ago. Mostly for the same reasons that I mean, the rest of the world was in terms of privacy and so on, so I’m not… I’m kind of hiding myself in this 900,000 large record collection for a few years. But sometimes I am on Twitter. So you’re certainly welcome to follow me there. My handle is @thdose.
Jorge: Fantastic. Well, I’ll link that in the show notes. Thomas. It’s been a real pleasure talking with you. Thank you so much.
Thomas: Thank you Jorge.