How Deezer is Approaching AI Music And Payments

0:00Roughly, if we look at the catalog being uploaded every day on any streaming platform, we are talking about a bit more than a million a week. And so half of that now is 100 % generative AI. So roughly 90 ,000 per day is generative AI. And it could be more than that, but it's 100%. 90 ,000 songs are uploaded every day. uploaded every day that are 100 % AI generated. Exactly.

0:39What's going on? Welcome to The New Music Business. I'm your host, Ari Herstan, author of How to Make It In, The New Music Business. The book, third edition, is out now. All platforms, hardcover, ebook, audiobook, however you like books, you can find the book. Today, my guest is the CEO of the streaming platform Deezer, Alexi Lontanien. And he is coming to me from France. Now, Deezer is based in France, and they're a very prominent streaming service in France. That's their largest subscriber base is based in France. They have a lot of subscribers in Brazil as well. But they are in 180 countries around the world.

1:17They have around 9 million total subscribers, paying subscribers, and about double that of total users. But Deezer, most notably, and maybe how you've heard about Deezer, if you're not a current user of them, they've been making headlines recently because they have been leading the front on AI music detection. They're the ones that are regularly coming out with how much music is AI generated on their platform. And, you know, the biggest headline most recently is that 50 % of all music uploaded to Deezer, distributed to Deezer is 100 % AI generated music. AI generated, it's crazy. And so, you know, most distributors and labels send the music to all the streaming platforms, Spotify, Apple Music, Deezer, everybody, Amazon Music, you know, all of them.

2:07So it's safe to say that it's probably about 50 % of all music distributed everywhere is AI generated these days, which is kind of crazy and scary. So we talk a lot about this. We talk about their policies around how Deezer at least is approaching AI music. They have, I would argue, been kind of leading this front. You know, Spotify has some policies around it. Apple Music a little bit. Amazon not so much. I mean, every streaming service is kind of updating and evolving their policies every day. You know, Tidal is demonetizing AI-generated music. Bandcamp has banned AI music. You know, I had the CEO of SoundCloud on not too long ago, and they discussed how they're approaching AI music.

2:52So I'd encourage you to go listen to that episode as well. But this was really fascinating. And we talk about the payment method, how Deezer pays out in general, how they pay, which was fascinating. We really got into the weeds of payment and fairness around payment. We talk about it was really fascinating to hear his approach and what he thinks is fair in terms of payment. Some of the types of deals that Deezer is striking with labels and distributors and just how they are paying out royalties based on the streams, how they are approaching, paying and designating AI music, what they define as AI music.

3:30So that was really fascinating. All about, you know, fraud detection. So he gave some really interesting statistics of the amount of AI music, which I would guess you might imagine are fraudulent streams and just like where these fraudulent streams come from. And I asked them to identify and define what fraud actually means and how they define what fraud means, because that's really important, how all of this works. So we had a really fascinating conversation just all about where we're at as an industry right now with AI music, with royalties, with streaming, and where we're heading. So this was a great conversation.

4:06Regardless if you use Deezer or not, you're going to get some tremendous insight into how us as an industry is approaching this. And, you know, he's in the rooms with all of the labels having these conversations and working out these deals and these licensing agreements and all of that because all the labels and distributors are sending their music to Deezer just like they're sending it to Spotify, just like they're sending Apple. And it's safe to assume that these agreements are similar across all streaming services and just how this is happening. But we are evolving as an industry. We, you know, streaming has been around for 15 years or so, and we are in a very different world right now than we were when streaming originally launched.

4:48So definitely think you're going to really dig this conversation. You can read more about what Deezer is doing at newsroom.deezer.com. This is where they publish all of their reports and everything. All the biggest trades have covered their press releases and everything there. I appreciate Alexi coming on the show and talking this through with me. I threw many hard questions at him and I appreciate that he did not dodge them at all and he addressed them head on so find all of us that make the show happen at Ari's Take on the socials you can find me at Ari Herstand on Instagram visit Ari's Take dot com get on the email list that's actually the best way to stay in touch with us is get on our email list but if you wouldn't mind right now just pause this episode leave us a five star review however you're listening to this show wherever you're listening to it give us that five star review leave us a comment if you're able to leave us a comment I love reading the comments Hit the thumbs up button if you're on YouTube, hit the subscribe button, hit the follow button, all of that good stuff.

5:42Stay in touch with us. All right, let's kick into the show. Alexei, welcome to the show. Thank you for having me, Ari. Absolutely. So, you know, Deezer has been making a lot of headlines over this last year because of your focus on AI music detection. I would say that you have been leading this front in the AI music detection and ahead of all the other streamers, all of your other competitors. I think you have taken kind of the most aggressive approach to detection and acknowledgement of, I would call it the issue plaguing our industry. That is my words. you know I don't make light of that I'm no fan of generative AI music and that being said I know it's out there I know it exists I know there's a lot of gray which I do want to get into and discuss it with you but before we get into the nitty-gritty of all of it for those that have not been following the news as closely as I have I would love for you to just give us an overview you of where we are at as an industry when it comes to AI music and what your research and what Deezer has been doing in kind of on the detection front and just your policies, I guess, of how you are approaching AI music in, you know, where we're at in September, October, 2026, as I'm sure things will evolve.

7:18Okay. Maybe let me start with the basic, which is for those who maybe have lived in a cave for the past two years there is now a lot of ai music that sounds pretty good like 97 of the people that we surveyed were not capable to identify that ai music were ai music and couldn't differentiate it from human-made music so that's the big element i think it started two years two years and a half ago. We started to detect that middle of 2024. We have been, Deezer has been always very forefront in terms of catching fraud, which is people basically trying to get reality away from the streaming services and basically generating AI music since 2024 has been one of the new tactics to do that.

8:10And so that's kind of how we started Deezer to identify those. So we saw this is AI music coming in vast amount. You were asking about where we are at right now. It's pretty massive. Roughly, if we look at the catalog being uploaded every day on any streaming platform, we are talking about a bit more than a million a week. And so half of that now is 100 % generative AI. So roughly 90 ,000 per day is generative AI. uh and let me just clarify more than that but it's 100 percent 90 000 songs 90 000 songs per day uploaded every day that are 100 percent ai generated exactly exactly straight out i just wanted to let that land because that's wild that's okay that's crazy keep going so that's kind of the layout the land today it's evolving fast because when we started to detect and communicate about those numbers it was January 2025 and we were more around 10 % or 15 % so we're talking about you know 10 ,000 15 ,000 songs the overall amount of songs keep growing by the way and so it has evolved in a year and a half it has evolved to now 50 % so I think the weight of percentage is increasing slower now but it's you know it keeps increasing so let's see where it goes.

9:39Talking about policies. Sorry. Before we get into the policies, real quick. I don't know if you just mentioned this or not, but I've read this. Of the 90 ,000 songs being uploaded every day that are 100 % AI generated, what percentage of those songs is that of your total amount that's being uploaded of all the songs? So yeah, how many songs are being uploaded that are not AI generated i guess what is the percentage of that breakdown yeah so roughly now 50 percent of all the song being uploaded every day are 100 percent generative ai got it crazy okay um so yes please tell me your policies on how you're approaching all of this so obviously you know difficult question um what do you do with those um because we are obviously a platform made for the artists we are our dna to support the music industry and the artists that that makes it uh interesting at the same time we are also a platform for the user that you know sometimes wants to listen to some ai song that are actually good quality or they actually like and so the the stance we took was to say okay we want to be fighting against fraud i think that's one very clear theme that we have and as i was saying those ai generative songs are massively used for fraud so up to sometimes 85 % of the stream on AI song are for fraud.

11:07So that's an obvious stance. It's very, you know, binary. We see fraud behavior. Those stream are excluded from, you know, counting in the royalty pools, obviously. And those songs that have more than 30 % of fraud on AI, they are just, you know, removed from the catalog. So that's, I would say, simple policy. You said the songs that you're noticing, sorry, what was the 30 % number you said that used for fraud? So if you have a high fraud rate on your song, an AI song, more than 30 % of your stream are identified as fraud. In that case, we just remove those songs from the catalog. 30%. Okay. And what do you deem as fraud?

11:55What is fraud? Fraud is a very complex topic that we're working on for a very long time. It's basically weird behavior that we see. Weird behavior that we see as, you know, the user is behaving weirdly, is listening at weird times with weird loops. We spend our time, you know, analyzing what a user wants to listen to. And so it's very easy for us, obviously with machine learning, I start people looking at it, to see any behavior that are non-human. So it's both behavior, basically. That's what enabled us to identify fraud. anything that is bought, basically. So that allows us to identify the AI songs that are used for fraud.

12:38And so those AI songs are just removed from the catalog. For the rest, it's a bit more tricky because as usual, there is some good quality songs that you don't want to remove. And so the policy we had was more a policy for the user. And so the user, what we know now, is that the user wants to have transparency. They are not against AI music. what they want is to know when they are listening to AI music and the use case is obvious I remember C &R Rose was a big example where a lot of users got really attached to the artists and the songs and then when they discovered it was a full AI project they got actually very frustrated and so that is what is important for us is to make sure that we don't push youth recommendation AI songs so all the AI songs are removed from recommendation and all the albums that have significant share of ai song we are tagging them so that you know if you start you know to listen to an album an artist and song you will know that these artists those albums are ai when you say that the songs the ai songs are removed from recommendation is that both edit like human editorial playlist that your internal uh staff will curate and also algorithmically generated playlists or even like algorithmically recommended songs?

14:03Are they just removed from that entire ecosystem of recommendation? So first is the automatic recommendation algorithm that by default, all the AI songs are removed from them. The manual recommendation from our editor are obviously manual. So they know if they want to put an AI song, I don't think it happened yet. They theoretically could If they think that it makes sense in the playlist, they could manually put it, but it would be a conscious decision for a good reason. Okay. Okay. No, but that's helpful to know. So just to clarify, if someone is listening to just, yeah, and a kind of an automated algorithmically generated playlist and they're just, you know, let the algorithm send them their music and maybe it's an artist radio station or something like that or whatever, it will, they know they can at least feel safe that they're not going to be fed songs that are 100 % AI generated.

15:06Exactly. Cool. um okay great what else in terms of how you're approaching this what are some other policies that you have i think that's about it i think it's also communication i think that was actually very important you know we are a big uh streaming player in france in brazil but in some part of the world we're not as big as some other and so the fact that these are actually made all those numbers known was actually it's not a policy but it was actually something very important for everybody to understand the state of the situation and how it's evolving and one number we discussed about the catalog being uploaded the question is also obviously what is streamed and is it generating any income and taking income away from the real artist obviously with all the policies that we have in place basically you we see three percent three four percent of stream that are going to ai song out of which a lot of it is fraud so that we demonetize but in the end you have you have 0.5 of the stream that are on ai song people are genuine people are actually looking for those songs and listening to them and so those songs are generating some income which are roughly 0.5 % of the total royalty pool.

16:28Got it. Okay. So that is interesting. So even though half of all the music being uploaded to Deezer every day, only 3 % to 4 % are, it's only generating 3 % to 4 % of the total number of streams on the Deezer platform in total. um now again the three four percent again it's it's a lot of it is fraud so the actual real people listening and actually monetize ai streaming is 0.5 percent is a half a percent i got matters right for the industry that's what matters so so of the that's i mean that is interesting so in the time period because i'm assuming you pay out what on a net 30 basis net 90 something like that um it's you uh from the time that they're getting the streams uh those three to four percent you have detected and then that that many of those streams are fraudulent so then you have demonetized those streams and in the end only a half a percent of uh all these streams are monetizable and that are being paid out to those uh creators or whatever you want to Exactly.

17:44That's very important to know, right? Because we say... You're not going to catch me call them artists, by the way. Okay, got it. Now, so that is, you know, I guess it's your combination of your fraud detection tools and your AI detection tools. um now i'm curious if you have a policy or you've thought about this um in terms of demonetizing generative ai music or at least making a different category for ai music that is monetized at a different rate so it's not necessarily competing with uh you know human generated music or how you're even approaching just the payout methods for yeah all these streams in general yeah i think it's a more difficult question so everything that is fraud is very straightforward you say okay it's fraud right you demonetize you remove them it's perfect right then you have all this gray area which is let's call them creator here uh but you know some people could argue they're artists of people spending time and using generative ai tool to add some you know instrument to their song to finish the production of their song.

18:58And they are uploading those songs to our platform and they are like, you know, why is that so different than any tool that we're using in the past to help create music? And so why should I earn from a different pool if the user actually likes those songs, actually looking for those songs? So we make it a bit harder for them, but once they break it and manage to get user and fans, why not paying them, you know? well i mean i would i would push back on that argument a little bit as people saying it's a tool how is this different from any other tool that's using like a drum machine or a synthesizer there's the argument that people have used in the past is like people are having these same arguments in the 80s when it was coming to drum machines and since and all that stuff um now there's a difference here because we've been talking about 100 percent ai generated music So the difference is, and this is a fair argument and a fair discussion to have with some of these producers and artists out there who are genuine producers and artists that might need a bass line or something.

20:04and they can't play the bass and none of their session musicians are able to come over today. They got to finish this song on a deadline. So they turn to a generative AI music creation platform, say, make me a bass line for this, or even in their DAW, because some DAWs, there are ways that you can prompt bass creation these days and loops and all that stuff. And so then they generate a bass line through a prompt. you know they don't play it on a keyboard like you know with midi or whatever and it's prompt based and so that is generative ai but now it's just a baseline now should that be tagged as ai music now i'm assuming in your uh cases that's a no because it's not 100 maybe the whole rest of the song is human created a producer makes it you know a human sings it all of that is just the baseline so you know sure i would argue that that can be part of the pool but it is great because like how much of it then maybe what if it's like uh an ai generated voice but it is a human generated production how do you approach that um i mean that that's a very difficult question uh that a lot of people are you know asking themselves and i think a lot of people are trying to align on this but again back to what we're trying to achieve which is being a streaming platform that connects people that creates and and fans that want to listen if someone has put a lot of energy in creating song in creating the lyrics in spending time back and forth with a tool that is very powerful tool i agree and very different from before i'm not saying the same it's different but it has put energy to create something that then people are actively willing to listen to there should be a monetization for them.

21:57And it's maybe something that at some point we could decide is a bit different in terms of monetization. I don't disagree with that. And I think we need a broad alignment. We are here to do a baby step. We started with some simple policy. We can probably go beyond that. I think it will be fascinating discussion with typically the Gen AI tool to be able to define how much work has been put in there. And so where should the money go? Should it go more to the catalog that has been used to train the AI? Or should it go a bit more to the person? Because they actually spend a lot of time to modify it.

22:30I think that's the future of the discussion that's going to be fascinating. Right. And, you know, now we've seen recently Suno's new model that is the first model that is version 6 that they call it has been created in collaboration with their music partners. It's just a way, a legal way of them essentially saying that they are now finally licensed some of the music, the catalogs of Warner, of BMG, of Believe to train this new model. Whereas the previous five models of Suno were trained on, I would say, stolen music, on music that was not licensed. and so you know that was the discussion there but now hopefully moving forward if they are going and all the generative ai you know soon is just the biggest but um hopefully all the generative uh music ai platforms will license from the rights holders and they will also be able to tag uh like you said and know how much of this creation was based on the training of various artists and of course it's a it's a clear you know remixes are a big thing and i'm curious how you're kind of you're approaching remixes but uh you know we re i recently had a label on the show who's they discovered uh an unauthorized ai remix of one of their artist songs this is a stick figure song and they talked about you know it was a fully ai generated song of uh one of their artist songs and And so one could argue it was a cover, but then it was they kind of used a lot of the they, you know, they thought that they were used some of the sound recording, but it was clearly prompted based on this song.

24:20But there's just no mechanism right now to get the original artists paid for this remix that was unauthorized. So they just ended up ripping it down. But yeah, there should be. We are at the baby stages of this. And so that is the discussion is just like, how can that money get paid out and broken up in a fair way? And to be fair to the music industry at large, I think it's making quite good progress and much faster than if you look at the last disruption of, you know, streaming and piracy and so on, where it took a lot of time. I think here they are very proactive. I mean, if you look at the most famous company, Suno, I think they already have a deal with Warner, Believe, B &G.

25:03It's moving quite fast. And so we are going in the right direction. They're also being sued by everybody else. So it's like both sides here. Right, right. Yeah. But yeah, we are obviously in favor of a deal being made. I think those two are here to stay. The best is, of course, that they remunerate the catalog that they are training on. And so then you will have two setups. You will have the original creation. So a creator that I would call an artist is spending time creating something, using those tools, but creating his own song. So here you will have a debate on how much should go back to the catalog, how much will the artist or creator keep.

25:49and then you have another story which takes a bit more time to to to exist but we are in active discussion i think everybody's industry is active discussion to create that which is the whole remix which is you start from an existing song and it's very clearly tied to an existing song or maybe two songs or three songs and then you create some new experience uh where it will be much easier to you know have most if not all the money going to the original artist publisher and label and and also the experience would be much more different because then you will you know create a really a fan to artist relationship i think that's what is very interesting in the remix side is that we will find finally finally there's a lot of solution but like you know on streaming application there is not that much interaction that much thing that the fan can do to interact uh you can live you can play this but then you know remixing will be a fantastic solution to be able to co-create if i may uh you know participate sure the creative process of an artist that you love are you having uh are you planning to implement a remix tool directly within Deezer?

27:10So actually, we already started to implement a remix tool within Deezer. We started in July. We have a non-AI remix tool for now because the AI discussion, a lot of sensitivity that we completely understand. But having a good AI tool that is fully licensed is still a challenge. Discussions are ongoing, but there is not yet an AI tool. It's coming. There is some good AI tools that are coming, but it's not fully bulletproof and perfect quality. But we are in discussion with a few of them. And we're also in discussion with the label to define how we'll remunerate, who's get what, which is obviously a very important discussion.

27:52It takes a bit of time. And so to accelerate things and test the user appetite and the UX and how this will all play out, we actually launched a remix solution with a few songs. So we have, I think, 80 now that we tested. So we have a 10 every two weeks song that we propose to fans to remix in a non-AI way for now. So you have, but you can still create 2 ,500 different options in a very nice and, you know, like the old remix way. You know, you have a lot of different stamps that you can mix together, a lot of effect you can apply. And we see a super good traction. User love it. Mostly young user, without surprise.

28:37And it enables us to understand which type of song works better, which type of artist, what type of user, what do they want. And that will enable us to go faster once we have the agreement and the tools to do AI remixes. So these 80 songs that you've made available for users to remix, where do they come from? Which labels and distributors? Everyone. I think everybody was very excited to test. and we have big artists. It's a bit French oriented. So we have artists like Céline Dion that you may know and then a very famous French artist as well from all type of labels. Actually, they are pretty impressive in their willingness to test and learn and move at the music industry speed, I would say, in the new solutions.

29:32Cool. And what is the payment method currently right now? What is that breakdown? Is all the money from those streams, are they getting split up amongst the Remixer creator and the original rights holder? Or is it just all going to the original rights holders of the label, essentially? 100 % to the rights holder. Got it. Okay. Yeah. Yeah, and that is something that I would imagine will evolve as this becomes bigger business. Similar to how YouTube has created these distinctions between the uploader, the creator of the video, and then if there's music in there, there's, of course, royalties that are paid to the master rights holder.

30:17That's typically the label. To the publishing rights holder, that's typically the publisher and the songwriters. and then of course there's maybe some ad revenue that's generated just to the creator the uploader of that video when we're talking on youtube and there's you know a few other little things of how they differentiate in various uh revenue streams and so i'm curious and imagine when it comes in the future as it evolves maybe the the revenue streams will become differentiated uh Maybe there will be a little bit that will go towards a remix or something like that in the future. It may. I think from our perspective, for now, it's really a fan tool.

30:58It's like, I'm a fan. I want to play with the music. I'm not here to make money. I'm here to potentially connect with the community to eventually being recognized by the artist. If I do something amazing, I think that for now, that's more what we think. If one remix breaks through, maybe, you know, they will have some specific arrangement. but we are more focusing right now on the experience and how we can gamify this to make it super fun more than rewarding monetary the remix which is great which i i love that and and you know exactly fans are not doing this to make money they're doing this to have fun and and to enjoy it and and honestly you know some of these other uh remixes like the the remixer the person that made this uh unauthorized remix of this stick figure song that blew up and it got you know millions and millions and millions of streams um it was just like some young kid from somewhere i don't even remember where who wasn't doing it to make money literally the song just happened to explode and catch on tiktok and then just catch on you know the streaming platforms um but so they weren't in it for the money they were just in it for the love of it and then it became a business because all the money was going to this kid and then a label swooped in and it was know and it became a whole mess so it would be helpful if there was some mechanism i'm curious like when it comes to remixes that are created off platform uh off deezer and then distributed to deezer i'm sure you're seeing some of this these unauthorized like ai remixes are you kind of approaching that agnostically and at this point at least you're just like if if a label issues us a takedown notice or a copyright infringement, we'll apply it.

32:39But until then, we don't really have a mechanism of how we're paying out differently or anything like that. No, no. And we had a few cases like this. So as soon as it's detected, it's taken down. And then one of them interested the label re-signed the creation and republished it and actually monetized it because the artist was OK with them. They made a deal with the initial creator. And it was actually a good hit. So it can happen, let it end up positively. Yeah, that's smart. Now, I think, you know, the reason that a lot of this, reason this is an issue when it comes to monetization and payment and how the money flows, I think is the way that Deezer calculates royalties.

33:26Can you just break down how you pay just from a very macro zoomed out level? how does Deezer pay for streams? So I will start with how other people pay and how it was originally done because that's simpler. And then I will explain where we're going and what step we have taken. So the original mechanism, because I think it was just simple, was we take all the stream, we take all the subscription money, 70 % of that, and we split pro rata to the stream. That's what's called the market-centric or pro rata. That was the original payment system. The problem with that fundamentally is if your audience is younger, basically you are taking away money from artists whose audience is older.

34:14If you compare that to the previous system, which was the CD in the previous system, you were buying a CD. No matter if you said it to a young person that's going to listen to it 10 ,000 times, or you say to another, another type of person that listened to it 10 times, you pay the same thing, right? on streaming, it's completely different. If your audience listens way more than another artist audience, you basically take money from that other audience for you. Which I think by any means is quite unfair. Like if you look at from the user perspective, it's like, I'm paying my subscription. I'm expecting my subscription to go to the artist I listen to.

34:51Which is not the case in the market centric. And I think that's where it's not a good approach. we have also of course other issues which are you know fraud and so on that we want to reduce we also want to value artists that manage to get users to search for their song proactively yeah and so that's where for all those reasons we want to move to this more fair system but the music industry is quite complex there is a lot of stakeholder each time you move something it has a lot of financial impact for a lot of people so you have to be a bit careful so the The way we approached it started in 2022 was to introduce the artist-centric model in between basically the market-centric, which I described before, and the user-centric, which is basically the money of one user is going to the artist he's listening to.

35:45And so we created this in-between artist-centric, which is one, capping the impact that one single user can have on the total pool. So by capping it to a thousand songs and 500 songs, then I did it to, you know, to one so that everybody has the same weight. That is a way to do it step by step to not, you know, shuffle the whole money around too fast. And so to help people to adapt to the new system step by step. So that was one of the things we did. We are basically in the middle of the gate. The objective is user centric. We've been very clear on that. but to manage the different stakeholder we're doing it step by step and I would say right now we are basically in the middle of the gate in terms of impact and the good thing that on top of that what we introduced on artist centric was removing all the noise so on top of the AI discussion we have also removed all the songs that are washing machine dropped on the roof all those kind of things we just removed them from monetization we recreated some something like that.

36:54Boosting the stream that are searched for. So if users are searching for things, it's active stream, what we call. They can't double. And also to reduce the impact of fraud, we can't double the stream from artists that have more than a thousand stream per month. And that is very positive to get rid of the fraud that we don't catch and bring back more money to the real artist. So that's kind of an overall package, quite complex. I hope I didn't lose everyone yet. With the vision to go to user-centric, which is simple and fair, basically. Okay. And just let me repeat this or just clarify. There was one question I had.

37:38I didn't quite grasp your capping the individual user's effect on the whole royalty pool. you said it said something if one user was listening like 500 times or a thousand what is that number like if if i'm listening to a song you're saying like if i'm understanding this correctly i listened to one song a thousand times you're saying maybe it's only going to earn like half that or a quarter of the total monetization because let me try that and i'm really sorry we had to make that that complex again the vision is simple we are coming from market-centric where one user can basically grasp a lot of royalty and divert that royalty to the audience he's listening to and there is no cap if he listened to a million songs he will capture a way larger share than his friend who's listening to 10 song per month right and we're trying to move to a place where each user will attribute his own subscription to the artist he's listening to right and that's kind of the two side of the of the situation is a market centric is a user centric and what we are doing is doing step by step by capping the impact the max impact that one user can have and so basically a user is a capital depending on the label so we do that at the label at the stakeholder level so it's a label is a publisher and we do the negotiation one by one with each of them.

39:07It's either a thousand, either a 500, either one. Meaning at a thousand, after a thousand song, you're basically, and it's very hard to explain simply. It works from mass perspective. But, you know, after a thousand song, basically we keep distributing your money as if you had only listened to a thousand song. 500, 500. And one is basically user-centric. That's what it is. And so the objective is to move from infinite to a thousand to 500 to one, which basically means user-centric. Okay, I think I yeah, it is super complicated. And, you know, this artist centric model, which I'm going to push back on, because that is a that is a universal music group term that is a Lucian Grange term, I would argue it is not artist centric, if we're really I mean, it's almost like a George Orwellian doublespeak type of term.

39:58I don't think it's artist centric at all. I think that is their term. um so you know i know you guys have adopted that because you made an agreement with universal and with lucian and all of that and that was kind of where this whole thing came from so i do agree that user-centric payment model is the most fair and if i can just explain it just like i think repeat back to what what we think user-centric is if i listen to one artist this month my entire and I'm paying$15 for my dealer subscription, all$15 are going to go to that one artist less your 30 % commission. Is that correct? Exactly. That's exactly right.

40:40That's the fairest model, in my opinion. It makes the most sense. That's honestly how I think most users think their subscription money is working. Of course. That, you know, right. So, you know, and here's the other issue, which, correct me if I'm wrong, But one of the big issues with the pro rata payment model, which most streaming platforms, Spotify, Apple Music, you know, most of them, even title to an extent, all of them are functioning within this kind of pro rata model where it's the piece of the pie. and what they like to say is if your song gets 1 % of all streams on the platform, you will make 1 % of all the money that is generated from subscriptions and ad and less the platform's commission.

41:28Now, that's an easy explanation of what pro rata payment model is. However, that's also not accurate. Now, again, correct me if I'm wrong, Maybe it works differently from Deezer. And here's where this is where I would like to ask you the question. In the pro rata model or even this artist centric model or whatever, do you have different deals with every label? Meaning, if I distribute my song through DistroKid, and Taylor Swift distributes her song through Universal Music Group, and there is a listener that the same exact listener in the same exact day listens to one of my songs and then listens to a Taylor Swift song, will we earn the exact same amount based on our streams?

42:21or does Universal Music Group have a different licensing agreement and get paid differently based on their streams? Is it weighed differently than DistroKid or TuneCore or CD Baby or Amuse or literally any other independent distributor or label out there than Warner?

42:42So first, I agree with your definition of market-centric and the fact that user-centric is the key. And again, we're going there and we're in the middle. It's a bit hard to explain, but we are basically in the middle of the gate. I think on your point of being, I think the short answer is yes, but then the difference is very minimal. So overall, we distribute 70 % of the subscription. And yes, some catalog will manage to negotiate a slightly higher rate. and some catalogs who don't have, you know, the power of negotiation will have a slightly lower rate. We are talking a few percentage points, but yes.

43:27Okay. I mean, that's, so yeah, and that, you know, is significant. You know, a few percentage points might sound small, but that could be millions of dollars for the catalog when it comes to that. But sure. So, So, you know, that is that is just another reason why all of these models, these payment models, in my opinion, and I think most people's opinions, other than, you know, the the CEOs of some of these companies, think is totally unfair. There is no no reasonable justification that one song distributed by Universal Music Group should earn more money than a song distributed by any other distributor.

44:11like, you know, my song, Distributed Through District, it should not earn less than a Taylor Swift song. That just doesn't make sense. It literally does not compute with anyone. The reason that it's happening that way is because of, you know, the licensing and the pro rata model and all of this nonsense. So I do appreciate that you are pushing to move towards the user-centric. No, it's balance of power. I think it's based on balance of power. Yes, yes. And to be honest, you know, I have tried to get to the bottom of this and tried to understand what are all these licensing rates. Of course, every company is bound by NDAs and nobody can say, well, this is what Universal pays and this is what Districtit pays, this is what ToonCorp pays, this is what Warner pays.

44:48You know, their licensing agreements, and I'm assuming their licensing agreements are different with Deezer, with Apple Music, with Amazon, with Pandora, with Spotify. You know, every streamer is going to, you know, have a different negotiation, different licensing rate or whatever. I've gotten little glimpses into these licensing agreements here and there from various people and new streamers that have slipped and accidentally told me the rates and stuff like that. So, yeah, I think as an industry, I do appreciate that you are pushing to move towards user-centric and that you agree it is the most fair model.

45:22But the pushback on that, you know, if you look back at the balance, it's quite surprising how little difference there is, though. Because if you think about it, a streaming platform could not really exist without a Taylor Swift. you know it's very difficult for Deezer to say okay that's fine I won't have Taylor Swift on my catalog it's very hard while for some you know very small artists emergent artists you know Deezer could exist without some of them of course and so the balance of power when you think about it is extremely different and despite that the difference in payout is very little again it's a few percentage point so that's another way to look at it which is actually you know very reassuring for the music industry that doesn't work really 100 as a business i think that's what is positive about that way of thinking sure i yeah i mean and i guess it's just getting to the core of what is fair uh you know and and like taylor swift is a billionaire and and you know universal music group are is valued at you know tens of billions of dollars and it's like should we be just because they're so powerful giving all of our money there and then we as independent artists are begging for scraps down here you know please big big boss at universal music group please give us some scraps so we can you know it's like no i think it should be the opposite i think like they're already billionaires so let let them get half the money that we get because we need the money more like The independent artists actually need the money more to survive.

46:59Y 'all are already billionaires, but that's not how the world works. So I understand. And that's not capitalism. That's the only other conversation that we don't need to get into. But if you really want to talk about what's fair, I would argue what is most fair is to pay independent artists more than the major label artists because they need it more. They need the money more than the billionaires do. But that's just - I 100 % agree. And my point was not about fairness. My point was to say, surprisingly, the capitalism is not at full speed in this industry. That was my point. Sure. Okay. No, point taken.

47:31And I just, I do want to, we glossed over this a little bit, but part of the move is you have demonetized, I don't know the type of what you call it, but you've demonetized the kind of the functional sounds or the category of like you know raindrops on a wind sail like the sleep sounds or I guess the sound I don't know what you call this but what do you call it do you have a cat non-instrument song basically where there is no instrument we just recreate that and remove it yeah interesting and how do you like what about spoken word yeah like rap without music. Do you demonetize that or I don't know.

48:20I know we just removed the thing where there was no voice and no instrument and we said, okay, all that is clearly not the same level of quality. The stuff that people are putting on to meditate to, to sleep to, the functional stuff. I got it. Sometimes listen for hours and that's why it's kind of an issue. Got it, got it, got it. Okay. Okay, so you know, there is precedent to demonetize certain sounds that are certain songs that are on the Deezer platform. But again, if we get towards a user-centric model, if somebody is intentionally choosing to listen to an AI song and an AI creator, I personally have no problem if those songs are monetized because that user, all of their subscription money is going towards the ai songs that they have intentionally chosen to listen to like you said you know so it's like i think right it just even gets further to the point and and same with rain sounds like i also don't have a problem if we're in a user-centric model if uh somebody is intentionally wanting to listen to rain sounds uh and they are paying a Deezer subscription for rain sounds, like that their money goes to the rain sound creator.

49:38I'm okay with it. They have intentionally not chosen to listen to me, so I'm not getting any of their subscription money. That's fine. But the only reason it's an issue is because it is this like pot of money and you got to split it up. And like people that listen, like you said, all night long, you know, eight hours worth of rain sounds, that then heavily weighs that pie that way. and that it is taking money away from me, even though it shouldn't and it doesn't make sense. So anyway, very cool. Okay, so no, this has been super illuminating. Are there any, like, I guess, just globally, what else are you working on?

50:18What are some of the things you're rolling out? Like you mentioned the Remix tool, which is like super interesting, very fascinating. And that's pretty cool. Are you said you might be implementing like an AI Remix tool or something moving forward, depending on the rights or whatever. But like, yeah, tell me some other things that you're looking forward to and working on this coming year. Yeah, the other interesting thing that differentiates Deezer from the other platform is that we are also working for the others. We believe we have an amazing tech platform that is able to compete from a tech service standpoint.

50:53In some markets like, you know, France and Brazil, we are not as famous in other markets. And so in those other markets, we're actually happy to power existing music brands. So an interesting partnership that we just announced was with Winamp. I guess you remember Winamp. I remember the Winamp player. Yeah. Surprisingly, I still like 40 or 50 million users every month. I did not know that. No idea. I had no idea. And the number one thing that those users are asking, we know for 20 years, is to have actually a subscription business with music. And so we're going to do that with them, I think, in January or December.

51:37Cool. So we just announced it this week. So that's super fun. I'm super happy to be able to, you know, enable our technology to power other people that have, you know, some relationship with music and want to compete with the big guys. That's awesome. That's very cool. Well, Alexi, I really appreciate you taking the time to chat with me about all of this and get into the weeds on it and, you know, have these important conversations for our industry. And, you know, you have a purview where you are in the room having these discussions with the big guns and the little guns and all the guns out there as as you kind of, you know, think about these issues.

52:16And we are in a rapidly changing industry right now. and I also like I said at the top of the show I do appreciate your leadership on this and I appreciate that Deezer has been at the forefront of the discussions of and really pushing the conversation for AI music and I believe that because you have been so public about your efforts with detection and just thinking about how you're classifying and the ways that you are approaching AI music It has pushed the industry and your competitors and the other streamers to acknowledge it and approach it as well. So it is, yeah, I appreciate that. Thank you.

53:00We like to think as well, but thank you for saying that. No, no, and I notice it, you know, and I think the whole industry at large is noticing it and we appreciate it. Even though your footprint as a streamer is not what, you know, Spotify or Apple Music is necessarily, not yet. but it is I'd like to think that as well yeah but you are at the forefront of the conversation so I have one final question that I ask everyone who comes on the show and that is what does it mean to you to make it in the new music business what does making it mean to me making it it's for artists to be able to connect with their fans I think that's the initial promise of the streaming platform to make it much easier to break a lot of barrier that used to exist between artists and fans where they had to you know to print a physical thing to distribute it now you know next day you can be live on a streaming platform so that's done let's say but i think that was very important and so now our job is to make sure that you know they can also feed their family with that and so that's why we're so focused on you know fighting against fraud reducing the impact of ai music and so on so that most of the money stay with the artist ideally with a user-centric model one day so that it's even more fair and enable even more artists to make it in the music business.

54:20Love it. Alexei, thank you so much. That's great. Thank you, Ari. This episode was edited by Ruben Zarate with music by Brassroots District and produced by all the great people at Ari's Take.