The Music Business Buddy
A podcast that aims to educate and inspire music creators in their quest to achieving their goals by gaining a greater understanding of the business of music. A new episode is released each Wednesday and aims to offer clarity and insight into a range of subjects across the music industry, all through the lens of a music creator for the benefit of other music creators. The series includes soundbites and interviews with guests from all over the world together with commentary and clarity on a range of topics. The podcast is hosted by award winning music industry professional Jonny Amos.
Jonny Amos is a music producer with credits on a range of major and independent labels, a songwriter with chart success in Europe and Asia, a senior lecturer in both music creation and music business at BIMM University UK, director of The SongLab Ltd and the author of The Music Business for Music Creators.
www.jonnyamos.com
The Music Business Buddy
Episode 55: AI Vocals and Voice Swap with Declan McGlynn
Imagine transforming your voice into that of legendary vocalists like Robert Owens or Angie Brown with just a few clicks. The future isn't coming—it's already here, and it's raising profound questions about artist rights, compensation, and the very nature of creative identity.
In this extraordinary conversation with Declan McGlynn, Chief Creative Officer at VoiceSwap, we explore how ethical AI voice technology is revolutionizing music creation. Unlike many AI platforms that scrape data without permission, VoiceSwap builds all their models with explicit artist consent, using specially recorded training data, and pays artists a 50% split at the moment their voice is used—not months later when tracks might be released.
The implications stretch far beyond simple voice conversion. We dive into how AI voice models differ fundamentally from traditional sample packs (one offers finite samples, the other infinite derivatives), the challenges of valuing someone's voice in perpetuity, and the emergence of voice models as a new form of monetizable IP. Declan shares VoiceSwap's vision for a democratic marketplace where anyone could license their voice, potentially transforming how vocalists, producers, and even engineers collaborate in the digital age.
This conversation captures a pivotal moment in music technology where standards are being established that will shape creative careers for decades. Drawing parallels to previous innovations like auto-tune and VSTs, we explore how initial resistance gives way to revolutionary creative applications—and how VoiceSwap is working to ensure artists maintain control and receive fair compensation throughout this evolution.
Whether you're a vocalist curious about new income streams, a producer looking for innovative tools, or simply fascinated by how AI is reshaping creative industries, this episode offers vital insights into protecting your rights while embracing the extraordinary possibilities of this technology revolution.
https://www.voice-swap.ai
Websites
www.jonnyamos.com
https://themusicbusinessbuddy.buzzsprout.com
Instagram
https://www.instagram.com/themusicbusinessbuddypodcast/
https://www.instagram.com/jonny_amos/
Email
jonnyamos@me.com
The Music Business Buddy. The Music Business Buddy Hello everybody and a very warm welcome to you. You're listening to the Music Business Buddy with me, johnny Amos, podcasting out of Birmingham in England. I'm the author of the book the Music Business for Music Creators, available in hardback, paperback and e-book format. I'm a music creator with a variety of credits I'm a consultant, I'm an artist manager and a senior lecturer in both music business and music creation. Wherever you are, whatever you do, consider yourself welcome to this podcast and to a part of the community around it. I'm here to try and educate and inspire music creators from all over the world in their quest to achieving their goals by gaining a greater understanding of the business of music.
Speaker 1:Okay, so in this week's episode, we are back into the realm of AI tools for music creators as we shine a spotlight on voice swap. So if anybody has not heard of voice swap before, it's basically a platform that allows anyone to transform their singing voice using AI. They have an exclusive roster of artists that work in partnership with voice swap, who receive royalties for the use of their AI voice. So every AI model is created using training data that is recorded from scratch that is made specifically for the voice swap community and the voice swap platform. That's just scratching the surface. By the way, there's an awful lot to it which we'll go deep into in this interview.
Speaker 1:I also want to mention the man that I will be actually interviewing, and that is Declan McGlynn, who is the Chief Content Officer for VoiceSwap Now. Declan has had a very distinguished career as a leading music technology journalist. So he's worked for the likes of Rolling Stone DJ Magazine, google Resident Advisor. He's worked for the BBC Splice Native DJ Magazine, google Resident Advisor. He's worked for the BBC Splice Native Instruments, universal Music Group. He is an incredibly impressive individual. He sits largely now in the Createc area, so creative technology but of course, he brings into it a wealth of information from other sectors of the music business. So you know, it's no wonder they appointed him in that role. He's amazing, as you will see in the interview.
Speaker 1:So without further ado, I'm going to hand over to the interview with Declan. Guys, grab a pen, grab a paper. This is one of the most dynamic and insightful interviews I've ever had on the podcast. Ok, here we go, declan. Thank you for joining me. Welcome to the music business, buddy. I've been really looking forward to talking to you First and foremost, most importantly, how are you?
Speaker 2:I'm good. I'm very good. We were just talking about I'm based in Dubai. It is very hot and I mean it's my first Dubai summer, so I haven't experienced it. I've heard the horror stories when we moved here. It's like, oh, this summer's crazy. It's pretty intense. But you know, apart from that, the air cons are on. I'm all good.
Speaker 1:Yeah, well, you've got your air con. That's the most important thing mate. Indeed, yeah. So, declan, voiceswap is a fascinating platform. I know there's a lot more to you than just VoiceSwap, which we'll come on to in a bit as well. But just for the benefit of anybody that's not familiar with VoiceSwap or might be interested in using it but have never heard of it before, would you mind just sort of explaining the concept and the purpose of it? So VoiceSwap?
Speaker 2:is a platform and a company that lets anyone change their singing and spoken word voice into another voice using AI, and you can do that on our website. We also have a VST AU DAW plug-in so you can do it straight in your DAW and you can also do it via third-party API integrations. If you're a bigger artist, you want to implement it into an app or that kind of thing? Um, so they do. I guess you know there are many of these types of platforms around there. What makes us different is that all of the artists we work with we work with directly. Um they we sign them to our roster. All the models are built with their consent. We don't scrape any data from the internet. We don't take any data unauthorized data from anyone. Um, all of the training data for each model is recorded from scratch and we work with the artist to make sure that we're using the same microphone, the same room and the same audio interface, and it might seem like we're being picky with that regard, but actually it makes a very big difference for consistency in the training data. If you want to get the highest quality AI model output, it's really important that the training data is consistent and that the model can just focus on the attributes of that voice rather than, you know, potential compression artifacts or the room noise or different microphones, etc. So, yeah, that's a really big part of what we do. We really pride ourselves on the quality of our voice models, so we want them to be used in professional environments. Um, so yeah, that's kind of like the, the main ethos of the company.
Speaker 2:Um, all the trade, yeah, as I said, all the training data is created from scratch, which is really important. And also, um, we pay our artists the royalty at the point of inference. So that means that artists get paid a 50-50 split of the cost of a credit which I'll get into in a sec at the point of inference. And the point of inference means as soon as their voice is created. So if someone uploads a piece of a sample in Acapella, they choose their artists from the drop-down menu, like Robert Owens or Andy Barn, william Bailey, they hit convert and that artist gets paid at that point of conversion.
Speaker 2:Rather than some other platforms, they'll pay or pay the artist a royalty when that track is released on spotify, kind of like a traditional split. Uh, we could, we do that too. We can do that too, but, um, we uh, yeah, so we pay at the point of inference. Um, so that means that when a user comes to platform, they they can either subscribe to get a certain amount of credits per month based on their subscription to, or they can just buy one-off top-up credits and they can use those credits to convert voices and then, as I said, the artist gets paid a 50-50 split on the cost of that credit each time their voice is converted. So that's kind of like the overview.
Speaker 1:Wow, brilliant. So that's kind of like the overview. Wow, brilliant. Um, yeah, I noticed the um, the kind of the one-off credit system that people can buy into, which I thought was a really clever model actually, because it gives people opportunity to be able to try it without necessarily committing to it and then going, oh, actually, this is brilliant. I'm gonna just, you know, because you know, if you subscribe, like anything, you know you make a bit of a saving's, so clever that you decide to kind of have that point of remuneration straight off the bat, because that really that's a very important component for all this right, because you know if somebody's voice is used then you never quite know, you know where it might end up and I guess to a certain extent it kind of doesn't matter because the vocalist is being remunerated at that point of sale. That's been really really well thought through. Presumably that was part of the idea from the off.
Speaker 2:It was indeed. Yeah, the platform was co -founded by Dan Stein, who was a British producer called DJ Fresh had a couple of number ones in the UK in the 2010s.
Speaker 1:I didn't know that.
Speaker 2:yeah, yeah, yeah, dj fresh wow yeah, yeah, so he worked with rita aura and I missed dynamite on his records and he's obviously worked with kylie and the pet shop boys etc. So he's he came from that artist background very much thinking what would singers want from this platform? What will make them most comfortable? I mean, we started in 2023 and, and even you know, right now there's a lot of fear and a lot of trepidation about ai, much of which is justified, but back then it was a lot worse than that. There was everyone was just saying, you know, everyone was, they hear ai, they're thinking, you know sci-fi, terminator 2, etc. Apocalypse, um. And so dan did a lot of work to understand what art will make artists most comfortable to join the platform in the first place, and one of those was royalties, at the point of inference, essentially, and we kind of thought that was the right thing to do. Anyway, like you said, we want to make sure that, even if the song is never released, even if it's just for demo purposes or songwriting purposes or ideation purposes, the artist is still getting compensated because, at the end of the day, you're using their voice and that's their IP, and the art is still getting compensated because, at the end of the day, you're using their voice and that's their ip, and that's something that we need to protect. And we wanted to set a standard that says, um you know, because no, there's still, standards are still being set in the age of ai and rights, um, and the creative industries, creator tools. We wanted to set a standard that says, yes, there is a value in converting the voice, even if you never end up using that track for commercial reasons. Um, so we wanted to pay. That's why we wanted to introduce that concept. I should also say that we do so. The way that it works is once you convert the voice, um, you can use. So. Once you pay your royalty, at the point of inference, you can use that only for non-commercial terms.
Speaker 2:And we have two types of artists. There's featured artists, which is our kind of what we would call kind of A-lister artists, like Liam Bailey, angie Brown, robert Owens, folly Dot, master Funk Artists, who you know have a career behind them. And then we have our session vocalists, and while they're incredible singers, they're more of like traditional session musicians, and for those you don't need an extra license. So there's two different types of license. If you want, if you say, for example, you make a track using Robert Owen's voice, you're like this is great, I want to release this track. You then have to get a license from VoiceSwap and we let the artists set the pricing on their licenses.
Speaker 2:And so you go to the platform, you fill in the form, you upload your track with Robert's's voice on it, you select the amount of time, because there's different price tiers depending on how long the length of the sample is and then it gets sent to robert and then he will approve it and can't assign it um, and then you'll get that. Basically, you'll get like a pdf contract. That is your license, and then that's a buyout license so you can use that in a commercial way you want um, or if you use a session vocalist, you don't need to use the licensing. So there's two different tiers of licensing essentially, which is how you end up using a voice swap voice for a commercial project, um, so you can. So you pay your credit, you can use any non-commercial which includes social media, and then, once you want to release that, you have to get a license from the artist ah, gotcha, okay, um, it.
Speaker 1:Does the artist at that point, kind of you know, have any kind of control as to? Oh, I don't really like that song very much, I don't want to put my name to it, or is it just kind of they have to go with it?
Speaker 2:yeah, yeah, they can reject the license. Yeah, we allow them full, full control over that. Um, it rarely happens. To be honest, this is an unfeatured license. I should point out as well. This isn't. You're not allowed to say you know, johnny featuring robert owens? Um, it has to be an unfeatured license.
Speaker 2:Um, if you do want to have a feature, as in you want to say, featuring angie brown, featuring lane bailey, that's a custom negotiation with the artist. Um, you can send them the track and say I think this is a really cool track. I think it'd be really. You know, it's kind of in your, it's your vibe, it's similar to what you've released before. But I would love to release that as collaboration, either using your ai voice or, which has happened before, the artist loves what they've sent and they end up re-singing their ai cells for for real. Uh, so they get in the studio with that artist and then they re-sing the ai idea with their real voice and they release that as a feature then. So really, it's a way to, like, you know, fast track the demo process in a way where you can send it to the artist and say here it is in your voice, what do you think? So they don't have to. They can hear it immediately in their voice and see if they want to collaborate with you.
Speaker 1:Wow, that's really really well thought through. Okay, Do you know what, Declan? There was about three questions that came to mind, all of which you just answered. Oh no.
Speaker 1:Right, see you. Well done now. That's brilliant. No, it's really good because what, of course as you well know, what you're doing there is you're catering for a really, really wide user base of very different levels of music creators, which is brilliant because we're in a very artist-centred ecosystem now where a lot of people have the same tools, and that's brilliant.
Speaker 1:But what comes with that is a responsibility to use people's data correctly. Like, for example, I manage an EDM vocalist and she made you know she's got a lot of Spotify streams and very popular and whatnot. But she also made a sample pack which has been used quite a lot, but one very particular big hit and also many other people, and in that pack it says you know, you can't use her name and her likeness, and yet some people, most people, are actually very, very respectful of that, but there's a few people that aren't and they do tag her in releases on Spotify and actually Spotify do very little, if anything, to actually try and help with that, you know. So it's a difficult thing, which I'm sure will probably straighten itself out in in in the years to come, because it's still a fairly new thing, isn't it, to try and combat. But? But what you've done there is great because it means that you know there are going to be people that say do you know what? I just want to use one of these voices to represent this song to send to publishers to get cut with an artist.
Speaker 1:There's also, you know, there's a lot of media composers that I've spoken to about voice swap, because one of the things that they have to do is to. You know, they very often need a vocal on a song these days for background music, for television, and so, even if an editorial level, the vocals taken off is actually the vocal that sells it to them, even if they don't end up using it. So you know, there's a load of libraries out there that, like this, is perfect for us because we can kind of just cycle between different options and that's the perfect voice. But we don't need it to be signed off, you know. We just need it for a non-commercial purpose. So, yeah, that's extremely well thought through.
Speaker 2:Yeah it's just about it. It's about, I think ai is essentially an efficiency technology um. It's it's been given a lot of different grandiose terms and a lot of this comes from the tech sector, who are trying to inflate um, its capabilities, in the not necessarily early days, but the early days of mainstream ai um and because you know, for various reasons, that we won't get into around government contracts etc. But really we have to strip all that back. What we're trying to do is teach people that there's a couple of aspects to it. Well, number one is you're less likely to be afraid of something you understand. Fear comes from not understanding what's happening just generally in life. And if we can inform the artists about how the technology works, how the remuneration process could work a lot of which still has to be decided through collaboration with industry bodies, and but we are working hard on that as well and we can help them make it as an informed decision. And one of the things that brought me to this project was I used to write a lot about the impact of the streaming economy um and how artists were left out of those early negotiations um that end up ended up defining their careers and continues to define their careers to this day.
Speaker 2:And when I saw, you know, four years ago or even longer than that, when I first saw the glimmer of this type of technology and the breakthroughs that were happening around AI and the creative industries, I saw history repeating itself in terms of artists potentially being exploited, deals being done behind their back and then suddenly this is the new era, this is the way it is. Get used to it and you know not being able to have any power to change things. So when I first went down because I knew DJ Fresh from working with him on a few projects previously just around journalism and articles and for DJ Mag and things like that he called me because we chimed. We really chimed on like being both being big nerds but also both caring about artists and the industry and the creative industry being exploited by technology, which has happened many times in the past. So we just started to like brainstorm about what this might look like. What could we do. This is early 2023. So I'm going into the origin story now, even though you didn't ask me. No. This is fascinating.
Speaker 1:I didn't know this. Thanks, Dick.
Speaker 2:Yeah, so we just had like a three-hour phone conversation and I was like you know, we need to be able to protect the data, not just the outputs of the AI models, but how the inputs are being monitored to make sure people aren't scraping things. Obviously, you know, stem separation technology introduced a lot of complications around being able to track acapellas, because now everything's an acapella because of STEM separation technology. So that adds a lot of complications to being able to protect artists identities from unauthorized models. And but you know, we we wanted to show that we could still build very high quality models. We could do it in collaboration with the artists, we could do it with their consent and they could be paid for it.
Speaker 2:Um, and you know, it doesn't mean we have all the answers, absolutely not, and this is a moving target. I mean copy, like you mentioned. I mean, you know, even with sample packs, the artists that you're working with already has problems people tagging them, and this has nothing to do with ai. So the music industry is not a perfect place, um, by any means. And there is no silver bullet that will solve any of these issues, and what we are trying to do is say, okay, this is coming, this is here the pandora's box. And yes, you know we're not saying you have to get involved. What we're saying is, for the sake of your future career, you have to understand this technology and if you decide to opt out once you understand it, great. But we want to help you get to the point so that you're not being exploited and that there isn't a lot of fear mongering. Take the fear away and explain the process. And again, like I keep saying, if, after all, that the artist says it's not for me, great, that's fine. But do it from a place of knowledge. It kind of reminds me of you know, I'm old enough to have been through a few of these cycles and the same thing happened when BSTs first came out. The same thing happened when autotune first came out. I assume the same thing. But I know the same thing happened with drum machines and synthesizers first came out. Um, you know, there's a replacement narrative, that way that these things start autotune is going to replace all singers. Um, you know, vsts are going to. Well, vsts, I guess, is different, but sampling is going to replace all musicians and drum machines are going to replace all drummers. They're going to replace orchestras and of course there was disruption. I'm not saying that and that wasn't the case. There certainly was disruption, but what it did was allowed people to work faster, more creatively.
Speaker 2:Auto-tune breed, you know, breed it, breed it bred, bred, um, whole new genres and sounds. And, and you know, the whole trap was based on on auto-tune um, and still is, and and became a sound in in and of itself. And now auto-tune is used by every single singer in the world, but it doesn't mean they're using it to the extreme. They're using it as a tool to enhance their creativity or as an efficiency tool. And you know, back in the day when VSTs first came out, all the analog crew were like this is terrible. You know, they sound terrible. They sound really thin. To be fair, they did sound pretty bad when they first came out compared to how they sound now, but it's the same. It's the same narrative of, you know, pushing back, clinging on to what we have and then, slowly, over time, embracing and realizing that it's a tool you can work alongside and work alongside your existing workflow, your existing setup in your studio, um, and that that's the way that we see this happening as well.
Speaker 2:And also, I was talking to someone yesterday about this. But they were saying that you know, like all these things, like sample packs a great example. Sample packs is another thing. Some sample packs are going to replace every. Everyone, like everyone's just going to use sample packs. And now you know, half well, more than half, I'd say like 80 of producers use spice, but they don't. They're not dominate, their sound isn't dominated by that. They dive into it for inspiration, they dive into it to pick up little segments and bits and they pitch them down, they reverb them and they queue them and they make them their own.
Speaker 2:Um and but sample packs were really frowned upon back in the day, like massively. It was like really shameful to use sample packs. And even when it produced, like if a producer ended up doing a like a branded sample pack, it was always because either they were at the end of their career and it was like, oh, you know, they're doing a sample pack, it was a bit embarrassing or they just threw on, like you know, the dregs of their folder of crap that they had on their laptop. They didn't, there was no um consideration for it. And now doing it, like you know, it's doing a splice pack, a loop masters pack. It's like it's a really big part of an artist's career. It's almost part of their promo campaigns.
Speaker 2:So the attitude around the technology is always shifting and changing, and all of these tools auto-tune synthesizers, drum machines are efficiency tools, just like AI is an efficiency tool. So I don't see that I't see this technology specifically, um as being any different to those, but they all start off the same way with a lot of pushback, um. But the big elephant in the room around ai, of course, is how it's trained. That is. That is the difference when it comes to all those other technologies in the ai um, and that is something that needs to be addressed and it's being addressed very seriously by VoiceLab, but also by the UK government.
Speaker 2:There's been a lot of campaigns lobbying the UK government not to change the copyright law, which they actually went through parliament last week. We've been part of those campaigns, trying to push back against big tech companies being able and small tech companies being able and small tech companies being able to take people's ip without permission and train models on them and monetize those models without any attribution or any any remuneration of the original ip owners. Um, that's something that we feel very strongly about shouldn't be happening, and so, while we can get excited about the tech and we do get excited about the tech and when dan first called me nearly three, two, two years ago yeah, over two years ago we were excited about the tech. That's what brought us to it. You have an obligation to do it safely and to do it properly and to set some new standards, because the standards are being set for this new era as we speak.
Speaker 2:There's no real definitive lawsuit that's come to a head yet, so we don't know how this is all going to play out. But the more we can present real, tangible solutions that are in the market right now, that are operating, that are generating revenue safely for artists in the age of AI, the more we have something to push back with and say this is possible. This isn't a pipe dream. This isn't the creative industries complaining, throwing the toys out of the palm. This is a real problem and here's a solution. We really don't want to just complain. We want to say this is a problem, this shouldn't be happening. Here's five solutions. Pick one. And that's what didn't happen in the streaming era and that's what brought me to this project and to this place.
Speaker 1:And that the streaming era and that's what brought me to this project into this place, and that was a very long rant. No, no, no, that's brilliant, mate. I love it. It's fascinating because, you know, just going back there for a moment, you know it is like when the streaming services first started. It was like the major conglomerates just went into bat for everybody, um, yeah, and it, and, and what they set up, the whole 80 20. It's never changed since and it's going to be really hard to ever change it.
Speaker 1:So, you know, and I suppose in some ways the kind of the narrative, the defence line at that point was well, you know, this is better than what came before, which was the dark age of streaming. You know the dark age of piracy and BitTorrents, and so you know, at least we're trying to do something to change that which you know they have. But what you are talking about here is something which is very, very ethically sourced and ethically remunerated, and it sets a very good example for where we go right. You know, I'd be very interested to understand the actual technical part, like how many kind of like, how many songs or how many minutes of a of a human voice, uh, does your technology need to have in order to understand how to use that voice going forward?
Speaker 2:so we, we ask for 20 to 40 minutes of acapella audio.
Speaker 2:Um, you know that that probably is more than you is required, but it's really important for us to have varied range of recordings so that, um, you know, if we end up throwing away 20, 30 percent of it because for various reasons and we still have enough, so we ask for more than we need, just so we can get enough of a range, we ask people to perform in the widest range they possibly can. That represents their voice. So if they're you know, if they're like a musical theatre style singer, make sure that they hit those big high notes and the big epic moments as well as having very soft whisper style singing as well, if that's their style. Because the more information their model has about how your voice sounds at different volumes and different styles, the better it will be able to represent your voice when someone else comes to use your model. So you have to give it as much information as possible about how you sound. Um, so, yeah, we do that. We do that over 40 minutes as well.
Speaker 1:Okay, yeah, that makes a lot of sense, in fact. In many ways it's kind of comparable to sort of sample libraries like Contact or Spitfire, that kind of go. You know, you could just play three notes of a piano bong, bong, bong and we'll just synthesize the rest in a mapping zone. Or you can play that real piano multiple times in multiple different styles and we'll figure out how then that instrument becomes available.
Speaker 2:That's brilliant, I'm glad you brought up sample packs, because it's something that we've been talking about a lot on panels and at events with industry peers recently. Because one thing that's very challenging by AI voice models and will remain a challenge for AI generally is pricing. And one of the bug of bug bears that we have is when people compare AI voice models to a sample pack. So you know back in the day not back in the day, but like your artists that you manage. Let's say, you recorded a hundred samples for a sample pack and then you sold that. You know, put it on Spice or whatever you sold on your website, and that's essentially you creating that. You know, put it on Spice or whatever you sold on your website, and that's essentially you creating 100 master recordings. Licensing those master recordings out for a fee, that's fine, that's it done, but it's static. Those recordings are fixed. You can go on and continue to make more music not using those recordings and those that sample back doesn't really compete with you. In the same way, if you give someone an AI voice model, the AI voice model can create infinite derivatives. It can create infinite derivatives. It can create, like a create a hundred million samples using your voice If you want it to in theory. So you're not. It's not the same as a static sample. Someone could come in and create, you know, endless derivatives of your voice. And then it's like, okay, well, how do you price that? If you're giving away the endless iterations of your voice? What's the price point for that?
Speaker 2:I think, for simplicity, a lot of other companies have gone down the sample pack route and just said, okay, we're just kind of it, will price it in the same way you would a traditional sample. But the problem with that is, you know it's worth so much more than that because it's endless. It's it's not the same as having 100 samples, um. So, yeah, that that's something that we're pushing back on a little bit. But also, we, we don't have the answer for that either.
Speaker 2:I think it's something that if you price it to a point that you know, here's us here. Yes, you can get access to this incredible voice model, but it's 10 000 pounds, then it's never going to work because no one's going to buy it. But then you can't undercut the artist to the point that you're giving away endless derivatives for 10 quid. So it's like, how do you price someone's voice? And it's a question, it's a kind of philosophical question, because it's something we haven't had to face before, because until this point, rights were assigned to either a publishing or a master and they were fixed once it was recorded. Um, so yeah, it's, it's. It's a new challenge among many in the age of ai yes, it is.
Speaker 1:Wow, this is. Do you know it's so exciting to sit here in 2025 and talk to you about this because you know this was like you know, two, two and a half years ago you started talking to dj fresh about this and and to me it's very exciting as to where it then, how it looks five years from now. I mean, yeah, you know you mentioned there about, about the, uh, the packs, or whatever incarnation or name they begin to take in the future. Um, there is an opportunity there because there's just so many um, electronic music producers that kind of sit in. You know, I was talking, I had um, the ceo of youth music on recently matt griffiths, great guy and we were talking about the, the concept of loneliness amongst many young music creators and actually the idea that they um, you know, they wouldn't probably use voice swap because they don't want to sing something and then have it changed.
Speaker 1:They'd rather just work with a vocal. And so if some of the AI model voices were creating packs, sample packs, ie, people are being paid but they're kind of being bought out, like what Luke Masters do, right, they use ghost producers to kind of create packs and stuff, pay them out and then they sell it, that's great, that's wholesale business. Nothing fundamentally wrong with that whatsoever. But if people are doing that and then you're kind of using those AI voices on a generative level to use compositions, and then there's no IP right, there's no publishing, there's no master neighboring rights, anything like that, and you just sell fragments of that to people, it means that then voice swap, kind of as an entity, becomes more accessible to more people. You know, I mean there's just a world of opportunity with all this yeah, all of that, you know, I think we're going to.
Speaker 2:While while two years ago definitely was the wild west um as we were calling it then it's starting to mature a bit the space. Now people are starting to ask the right questions, um, in my opinion, around rights, um protecting rights, which is great to see, but there's still a long way to go and I do think we'll see that what you're describing happened, where a lot of sample packs, vocal sample packs, just pop up on various platforms and it turns out they were made using an ai voice model, not actually, um an artist, and and you know.
Speaker 1:Well, yeah, we'll have to see how that plays out. It's another issue. But, but, but also, you know the question. Then I know that there's a lot, of, a lot of what you've been fighting for, and myself as well, as you know fair rights and remuneration but I can't help but wonder if perhaps it's not always actually the answer, because, um, for it to be the right price point and accessibility to be there, and people can't afford a top end, and they go. I don't want to use that person's name, I don't want to tap into their data, I just want to use their voice, and actually what I want to do with it is something slightly different. I just want to use its source material to put it through a vocoder and create my own sound, and then, you know, then we're kind of getting into kind of the sampling side of it, which, but it all derives from having that original source material. So it's it's a very, very interesting, exciting space, and the answers are perhaps not always um in in, in how intellectual property is recognized or remunerated.
Speaker 2:Actually, yeah, that's one of the reasons we wanted to pay at the point of inference, because then, no matter what happens to the sample, even if it's edited out of you know you can't even identify it anymore the art is still getting paid. So, if you know, a lot of people do to this day sample and then try and obfuscate it to avoid paying any pay, like clearing the sample. That's an age-old thing to do, but even if they did that here, the artists would still get paid because they get paid at the point of inference anyway. Um, which is why we wanted to introduce that. So even if they take it, put it through vocoder, pitch it down, cover it in reverb, reverse it, re-record it, play it through a boombox in a tunnel in Brighton and record it whatever they want to do, like something creative, resample it 100 times, the artist still got paid at the point of inference, and that's why it's really important to have that there and try and set that standard for AI.
Speaker 1:Yeah, that's a really, really good point. Are there any? I don't know whether I'm kind of jumping the gun here, but is there like a text to voice option for corrections? You know, like if, if, let's say, I sing the word um there and it's been interpreted as bear, you know?
Speaker 2:can I? Is there a?
Speaker 1:I don't know why I think that example, but like? Is the technology got to a stage yet where it can be corrected by text or not?
Speaker 2:Yeah, for sure. I mean we don't have that feature, but we are working on text-to-speech. So right now we are voice-to-voice, which essentially means I pick up my microphone, I sing, I upload that acapella, I choose my artist, I convert it and it comes out in their voice. We are working on text-to-speech capabilities to allow you to do lyrics and spoken word, essentially with voiceover talent. Wow, a big part of our company is also the voiceover industry. We work with a company in Hollywood to do a lot of movie trailers and a lot of big branded voiceover work and we are currently creating their voice models to of big branded voiceover work and we're currently creating their voice models to be deployed for voiceover specifically. Um, so we don't just work in music, we work in in media and advertising and voiceover work as well okay, that makes a lot of sense.
Speaker 1:So, do you, do you work in the um the audiobook market as well? We certainly will.
Speaker 2:Okay, yeah, we, we are working towards all of those solutions that you can imagine, with text to speech and yeah right, okay.
Speaker 1:Okay, that's interesting because you know there's um, because I spoke to somebody because I had a book that came out last year declarer, and and I know that the the audio book rights were sold on it a few months ago because the publisher told me, um, I don't really have any involvement with it, they don't want me to do it. Well, they haven't asked me to do it and I'm not offended by that. But I just happened to kind of look into it and you know, a couple of people said to me oh, you know, I'd be a couple of voiceover. I said, you know, I'd love to, yeah, voice that book. I was like, okay, well, I'll put you forward, I can't promise anything. Blah, blah, blah, blah. But then I kind of stopped and I said you know what? Actually, you know, I'd love it if you did that. That's fantastic, but actually perhaps the more sustainable option for you is to actually have your voice available that anyone can pick from and you're kind of getting passive income for it, rather than sitting down and actually doing it.
Speaker 1:Now, you know, I know some people are really anti that kind of thing. I don't really understand why, because if you can't, it's like saying you know it's like someone composing an orchestra and going, oh wow, that's amazing. Oh, actually it was samples. It's like, well, it's still amazing, like I can't tell if it's a real person or if it's a because it is, it's not like it's a because it is, it's not like it is a real person. You know, it's like when you play a contact is when you go, that guitar sounds amazing. It's not a guitar, it is a guitar. Someone did play that. You know it's that and I think sometimes the education actually is there. That is is people are getting smarter to that kind of stuff. You mentioned earlier about vsts being a bit naff back in the day and all of a sudden they got better because of sampling techniques. You know, um, so that's really good.
Speaker 2:But also, you know, while in that moment, between vsts coming and or should we say digital audio tools generally, and the laptop and the computer being used and them becoming so good to the point that you couldn't tell the difference, there was an incredible amount of creativity when they were failing to do what they were designed to do. You think about, like apex, twin and square pusher and artists like that, who, who let they lent into the failings, in inverted commas, of that technology. Um, and you know, drum machines weren't designed to make house and techno, but they did because creative people got their hands on them and did. They weren't supposed to, they were trying to be I mean, not all of them, but most of them are trying to be real drum, drum, drum kits. But they failed because of the limitations of the tech. If you think about like the sp 1200 sampler, which is like a classic hip-hop sampler, and if they designed that today it wouldn't have that 12-bit crunch, because 12 they would have probably 32-bit float or something crazy. But people now they long for that sound, they want, they want it's same as the mpc 60, they want that crunch that that lower bit rate brings.
Speaker 2:But at the time that was a failing of that product in inverted commas again, failure, subjective, but so so I think that, as well, what we're going to see now is and also in theory, that kind of crazy autotune sound is a failing of the product too, because it's not natural, whereas autotune was trying to make it natural, but someone put it to extreme levels.
Speaker 2:Um, so you know, creative people get their hand on these tools, they do creative things and they create new sounds. I do think we'll have this like ai aesthetic, sonic aesthetic between even though the models are improving incredibly quickly between where we are now and the next couple of years, where there's like it's the models aren't good enough to be completely discernible, um, from the real thing, but in that moment there'll be something new. Instead, when the model falls short of being what it's trying to be, it becomes something different and something new that that didn't exist before, and there's been a whole bunch of art exhibitions of kind of failed in inverted commas uh, art that ai has created, but it's kind of beautiful in its own way because it's it's kind of, um, it feels very otherworldly and very bizarre to look at and it's kind of engaging. So who's to say that's not art? And I know that's a different podcast, so we won't get into that.
Speaker 1:But no, no, it's really nice to bring it out of music to give it a frame of reference.
Speaker 2:Sometimes that's a great example yeah, and like that and the same with synths. You know, since we're trying to be orchestra, they're trying to be orchestras. Um, if you load a string sound in a juno 106, it doesn't sound like a violin but it sounds really cool and interesting, and that was a big part of house music, techno and other genres. And so this technology doesn't just come to replace things. It brings with it something new a lot of the time, usually unintended, usually an unintended consequence of a lot of the time um are usually unintended using, an unintended consequence of a piece of technology coming along. Um, and we'll definitely see that with ai as well, and for me the most boring part of ai is making it sound exactly the same. It's like I don't want it to just create, like you know, like the kind of suno and udo style example of like create me a disco beat and it comes out as a disco beat. It's like that. That's not very interesting to me at all. It's like I'd rather it come out with something completely different that I couldn't even possibly imagine. To like inspire you, to drive you in a different direction. Yes, and you know, eventually it'll probably be both. But um, yeah, so there's a bit of a tangent.
Speaker 2:But to go back to another point you made which is really important is we also have a user models.
Speaker 2:So anyone cause you mentioned about monetizing the voiceover voice um, anyone can come to the platform and create their own voice model, um, and basically you can then use it inside your DAW.
Speaker 2:Once you make it on the platform, um, you can open the plugin and log in and you'll see your model there privately inside the plugin so you can continue to work with it in your DAW. You can also share it with other people. So if you have like a bandmate or a co-producer or even a label or whoever, if you're doing a writing camp or something like that, you can add some emails in the backend end of voice swap on the website and then when that person logs into the plugin, your voice model will show up in their plugin as well. Um, and then when, when the project's finished, you can revoke access because we didn't want people to just have your voice model forever, um, so say you're doing like a remix or something, or they want your voice model to do some ad libbing or whatever. You can send it to them and then, when they're finished remix, you can take it back from them and once you, once you delete them from your, from access to your voice model and the voice swap website, it'll disappear from their plugin as well.
Speaker 1:Whoa, I didn't know about that. Okay, yeah, and then the end goal, um.
Speaker 2:Also, there's another really important part to the user models as well, which is all the one thing. We wanted to launch user models. But we knew there was a risk of people just, you know, uploading Beyonce acapellas or uploading Taylor Swift acapellas, creating like unauthorized voice models of singers. So we work with a company called BMAT who are based in Barcelona and they are kind of like a rights attribution company, but they do a lot. They do something like I mean 13 billion checks a day across the internet and broadcast for data points of music that's being played on swedish radio or this youtube channel, and then they aside, they collect that and they assign those um rights to the relevant pros. And so we work with them to implement their technology into voice swap so that all the acapellas and all the training data gets uploaded to user model training. It's screened by BMAT to make sure there's no copyright content inside of it, because they have a database of 180 million fingerprints of audio and once it's confirmed that there's no match none of that uploaded training data matches that database then the training can continue and if there is a match, the training will get blocked. So people can't just come and upload any acapella. They scrape from the internet to create voice models and voice swap. So that means that those models are all compliant for licensing going forward.
Speaker 2:And so one of the things that we're looking at is licensing our voice models, not just our featured artists and our session vocalists, who are the artists that we've chosen, but also being able to license out the user models into other creative tool platforms. If somebody wants, you know, someone comes to us and says I need 15 female vocals models for this new plugin I'm building, or whatever. We can then go back to our users and say, okay, we've got this opportunity. Um, you know, do you want to be part of this? Users opt in or opt out of these opportunities when they first train their models. So you can say opt into licensing, um, and then we, when we get a request for licensing, we'll go through everyone who's opted in and then we'll be charged to the relevant models, depending on what the request is from the licensor, um, and then we'll say to them okay, they're paying this amount a year to license your voice into this product. Do you accept or not? So we're trying to find new revenue streams, not just for the established artists but also for, like you know what you might call the bedroom singers or the less established artists who still are incredibly talented but might not have the profile to be on the voice forab website.
Speaker 2:And then, eventually, what the end goal with all of that is? Because the models are compliant, copyright compliant. It means that we can go ahead and build a marketplace where anyone can come and monetize their voice and they can set their own terms. They can set their own price point. They can set their own licensing point if they want to have licensing at all. Their own terms, they can set their own price point. They can set their own licensing point if they want to have licensing at all.
Speaker 2:And eventually you can imagine having a section on VoiceHub where it's the marketplace. You go to the marketplace, you type in jazz vocalist, female, spanish, whatever and you get a hundred responses of those voices from around the world. You can work with them passively. You can either use their AI voice model. You can upload your track without the stems, just upload the full thing. We'll use stem separation technology to split it. We'll swap out the acapella with the model you've chosen and then you'll hear your idea in their voice before you commit to working with them. And then you can either pay to use their AI voice or you can contact them, say, okay, cool, this sounds good, let's get that recorded for real in the studio. Um, so that's kind of what we're working towards the moment good god.
Speaker 1:That that's. I'm kind of a bit blown away, declin. I'm so glad that I've got you here. I'm told this is amazing, mate, honestly. Um so do you have plans to, like you know, license your own kind of you know, set up your own label, have your own kind of entertainment agency as a result of this, as an extension of it, or is that you know?
Speaker 2:I think, totally. I think that that's seeing ourselves as an agency is something we have done since the start, like representing the AI identities of singers, essentially, and finding new monetize it, new ways for them to monetize those identities, be it through licensing into creator tools or licensing into new opportunities that we can't imagine yet that'll come out in the age of AI. I think that's what we want to do. We also want to protect. So if you're representing someone, you also have to be able to protect their voice. So we see creating a voice with us as like what we call a source of truth. So you come to our platform, you create a voice. Even if you don't make it public that voice, existing will then give you. It gives you some kind of legal ramification to be able to say this is the real version of my voice. Therefore, that one that I found on Google isn't the real one, so you have to take it down. So you need to be able to create a source of truth, kind of like the blue tick of your real voice, and then we'll represent your real voice, what we're working on. So we represent the source of truth of your AI voice model. Basically, while this takes a lot of collaboration across the industry. This is not possible today, but what we want to see eventually is the ability for everyone to collaborate together, which include DSPs and have some kind of enforceability, where we can say we have the contracts, the paperwork to say this is the contracts, the paperwork to say this is the real voice model of this person. So we can go and do a takedown notice across the internet on unauthorized voice models that are popping up left, right and center and to be able to find out when someone uploads something on a distributor or to a DSP like Spotify or any other platform or YouTube. And you know, in a dream world, in the utopia and of course the music industry is far from a utopia, so this is unlikely to happen in this particular way. But you know, you've got to aim for these things.
Speaker 2:I would love to be able to see a future where we have a database, which we have already, of voices that we own, voices that have been licensed to us, and then when something's uploaded to spotify or through a distributor, it there's a watermarking technology. It says this voice is was created using voice swap and then it pings our database to say if that artist gave permission for that track to be released in the first place and then if they say no, obviously it gets blocked. As you said, there are takedown issues on spot. Takedown things on Spotify are quite challenging at the moment, even for non-AI things. So we wanna kind of get ahead of that because we feel like there's gonna be a lot of take down requests in the age of AI and also it's our duty as custodians of your voice to go out there and police and the unauthorized use of your voice as well, so that the authorized one can continue to generate the revenue. So all of these types of policing things.
Speaker 2:There's a couple of startups that have come up recently. There's a couple of initiatives that are really interesting, but they're still again, it's all a moving target. It's all front of mind for us. You need to have enforceability, otherwise you can't protect. Simple as that. We can have your AI voice model, but if someone else can go and build it without your permission and we can't do anything about that, then you know it's not as powerful a message. So we know we need to go out there and police the voice models that could potentially pop up and any unauthorized contents being created with those unauthorized voice models that could potentially pop up and any unauthorized contents being created with those unauthorized voice models as well. Like I said, it will take a lot of collaboration and that might take a lot of time, but I think that's what we got to aim for.
Speaker 1:We have to aim for some kind of industry standard solution around protection of artists in the age of ar yeah, absolutely well, I mean, if you you a small-scale baby, example of that would be the success of, let's say, a YouTube content ID. You know that can just take a digital footprint and go right, we recognise that this is the recording of that song. It's just a more elaborate version of that, isn't it where we recognise that this is that person there, based upon these particular artefacts? Elaborate version of that? Isn't it where we we recognize that this is that person there, based upon these particular artifacts, and it might be difficult for us with the naked eye or with the ear to be able to look at that, but we already know that the technology exists in it's.
Speaker 1:But look at how spotify used raw file analysis. Right, they're able to hear things in music that most people can't speech ability, dance ability, rhythmic feel, suitability for curation, algatorial playlists, all that kind of stuff. It's amazing, it works. So the point is that the technology is semi there at least, if not wholly there. But it's the back end of it, it's the um, you know the legal side of it and and that partnership work, isn't it? It's almost like the world kind of needs to come together and go. You know all you guys there do you know about this. Yeah, this could change everything forever.
Speaker 2:Yeah, yeah, totally. And you know, youtube content idea is a brilliant example because that's the example I always use too of where rights holders were terrified of a new technology, where YouTube was getting sued into oblivion before Google bought it and then also after, when Google bought it, and they needed to find a solution where you could have the best of both worlds, where you and I could upload us dancing to Taylor Swift God forbid and we wouldn't get takedown notice. Well, we probably would, but the thing is the way the content idea works is a perfect world, because the artist decides what happens to the content. So anyone can go and upload. Let's probably. Let's say, you know, ariana grande track gets uploaded to youtube and ariana grande's team and ariana herself have decided that anything that gets uploaded to youtube in this region, we're going to monetize that, we're not going to issue a takedown. So then all that happens is they serve ads on that video and they monetize those ads and the video gets to stay up. So, you know, the 13 year old who uploaded it doesn't get punished, um, but the artist still gets paid and youtube can continue to have copyrighted content on there and everyone can win. You can, they can also block the content or just track it for analytics. So it gives the artist the power but the platform the freedom, and that's what I think is a great example of how um music, the music industry and the creative industries could work in the future of ai.
Speaker 2:It will take a lot more than that, because you know if you have just a song, you can issue that fingerprinting and then you fingerprinted it once, that's it. That's the process with, again, you have endless derivatives from an ai model. It's much you can't unless you fingerprint endless derivatives from an AI model. It's much you can't unless you fingerprint every single output of that model, which would be expensive for the company, or you fingerprint the training data, which is another interesting concept that could. But again, none of them are silver bullets. And also content ID is not a silver bullet. It doesn't always work. People get false copyright strikes, et cetera. So you know, it's just about being able to bring something to the table that could pacify all parties and get us moving in the right direction, instead of just saying, oh well, it's a wild west. Ai is going to. You know everyone's going to scrape everything, train whatever they want. There's nothing we can do, that's the end. You know, voiceover wouldn't exist if we thought that was the case. So we are very actively discussing with startups and other collaborators who want to solve this problem.
Speaker 2:But yeah, content id is kind of like the example um that I often use as well, because it does allow, you know, the individual to use copyrighted content safely while giving the rights holder all the control over how it's used. Um, but again, the thing is, youtube's one platform and it's owned by a very big company called google. Um, if you then open up, you know, spotify, deezer, apple, music, title, um, endless amount of dsps and then also the distribution, because youtube owns the distribution channel onto youtube, it makes a lot easier to do that as well, whereas if you're a dsp, you could use cd, baby, distrokid, tunecore, believe, blah, blah. So they all have to have the same solution integrated, otherwise it's not going to work. So that's where the challenge is yeah, that's that's.
Speaker 1:Uh, that's a. That's a good point. Um, so I'm just I can't help but, like, think ahead and start wondering what deals might look like in the 2030s. You know, can you imagine, you know? In fact, I might reach out to a couple of lawyers and ask them this actually, is that the idea of, let's say, you take an artist right and in 2032, they go and sign two completely different deals right, one for them as their artist with, let say, a distributor or a label, and then they do an ai contract and they're totally separate things?
Speaker 1:Now, right now, that seems really alien, but it's not beyond the conventions of possibility right now to actually imagine that that could be a thing. I mean, you know, ideally, with voice swap, they'd sign both to voice swap and that'd be great, right. But if there is a flexibility with an artist to be able to go, yeah, I'm gonna. This is my training and this is my own thing over here. People can do their own thing with my thing over there, but when I put my if you like, blue tick next to it, it's over here, and actually those two deals could be structured very, very, very differently. I know that there is that grey area in the middle, which is what you're referring to there. But could that happen in the future, do you think?
Speaker 2:Yeah, it's happening already. We know that labels are trying to sign AI likeness rights when they sign new artists. Now there's a very big debate and I'm not going to get into the finer details on it, but there's a very big debate at the moment around. Do labels already own the AR likeness through some of their marketing clauses? I can tell you, you know, without naming any names, that that's not the case. A lot of labels will argue that, you know, due to the fact that a lot of the contracts and, again, every context different, so it's very difficult to say this in the generals, with the general sweeping terms. But, um, a lot of the labels will say that because of the fact that labels, when they sign an artist, they own any content that's commercialized a lot of the time and therefore, if you, while they might not own the voice model, if the training data that was recorded to create that voice model then goes into something that is commercialized ie, voice swap or any other then they could have a say in the fact that they might own that product, in inverted commas. It's a very questionable argument and certainly hasn't been resolved, and lawyers I've spoken to about this have given very conflicting responses. And also every contract. You know it will come down to the wording almost, rather than it being a clause, and it'd be interesting to follow the progression of any lawsuits that come out of that. Because you know it's it's it's pretty dystopian to own someone's like, own someone's voice, um, you know outside of what they record, but just you know, you know they sing in the shower. You own that. It's kind of like, how far do you want to go with this? Um, and I think that that's a I wouldn't say it goes as far as to call it sinister, but I think it is a very worrying trend and I would advise any artists to very, very, very carefully read any contracts around which you will be doing anyway, but around AI clauses, because if you sign your AI likeness away in perpetuity, we don't know what this feature is going to look like and I think that would be a very dangerous thing to do. You know, also because it's a right-sparing asset after your death as well and it can be part of your state and we're probably going to see a lot more of that um, in fact, we definitely will see a lot more of that. We have seen it already and you know artists who pass away whose estate continue to release content using their ai identities. So who've passed away, whose estate continue to release content using their ai identities, so that's a monetizable asset too.
Speaker 2:So you have to be very, very, very careful when you think about the protection of your ai identity because, like I was saying before, it's very different to signing a catalogue of music to a label that they can monetize after your death, because that catalogue is static, it's a fixed recording, it is what it is. It's these amount of albums, it's these amount of songs, it's you know, these are the that's. It's got this value. You've sold it. The end. They can do whatever. If they have your voice and your voice model and your music model, they can just continue to make as much music as they want, forever and on forever.
Speaker 2:Um, and you have to. You have to rethink about different, the differentiating. Let me start that again. You have to think about how you differentiate the pricing around that like, how can you possibly put the value on your musical output in perpetuity, forever, that never ends, and infinite amount of it? Um, so I think artists and we we've been working with BIM a lot and we just did a four-week course with the students at BIM and week three was all about.
Speaker 2:This was really hammering home, protecting yourself, protecting your rights. Don't sign anything that you're not 100% confident won't come back to bite you in the future. It might seem exciting and rosy and a lot of people can get swept away by the idea of being signed and that's still a very exciting and romantic concept being signed but you have to be very, very careful, because your assets now can be used to build a model, without your permission, that can compete with you, and I think all of that stuff has to you, um, and I think we yeah, all of that stuff has to be front of center for all artists at the moment. That's you know. Let's go back right to start this conversation.
Speaker 2:The whole point of this was to educate artists to say this threat is here and this could potentially happen. And one of the things about ai and the headline most of the headlines around ai are very apocalyptic, very doomerist, and, again, some of that is valid. But what that means is that people just kind of panic and they don't think they're not thinking rationally about what's happening. And while they're over there panicking, the music industry is over here making deals and making decisions, like they did with streaming. So it's really important that artists get on top of this stuff and are very, very, very conscious of what they might be signing and what their assets are worth when those assets can be used to train, something that can create an endless amount of competing and derivative assets.
Speaker 1:Yeah, Declan, I've got to say you know I have a lot of like brilliant guests on this podcast, but my God, you are so impressive, man, honestly, like the things that you're thinking about here are about nine or ten steps ahead of so many other people in so many other sectors. What an asset you are. I mean, that's amazing, wow. I mean. I know you've come from a very exciting place. Your CV is amazing and you know, I encourage people to Google you and look at all the things you've done before. You did this and all the other amazing things you'll do next, I'm sure, but my God, I think if there were more and more people like you you do before you did this and all the other amazing things you'll do next, I'm sure, but my god, I think if there were more and more people like you, then, um, then, uh, I think the music industry would be even more exciting than it already is. Mate, I really do.
Speaker 2:Thanks, mate, that's wow. I'm stunned to silence for once in my life. Um, no, I think that my my whole background has always been about artists and and representing the interest of artists in the tech space. I always used to find it like when I because I used to work on future music magazine and then I went into content creation and then I did a lot of journalism with different magazines and consultancy around music tech for different brands like splice and title and google and rolling stone and stuff, and a lot of the time, like throughout throughout my career up until the past, probably since the pandemic, there was like the tech industry which was like over here, it was kind of like if you're at a house party, you have the musicians and the artists and the labels and the A&R team all in one, like 75% of the room would all be in one corner, like dancing and laughing and drinking beers, and the tech industry would be in this corner, kind of like snooty and nerdy and awkward, and that was the way the music industry was seen and I always had a real problem with that because I was like tech, like tech is culture, like every single person who's over on the other side of the room is using a diw, they're all using plugins, they're all using technology in their creative process.
Speaker 2:But then, when it came to discussing tech, and there was like the technical people and there was a non-technical people and obviously, you know, over 20 years, a smartphone has meant that, yeah, social media is culture now and it's kind of blurred into one. But I was always trying to say no, like these are one and the same. Like your creative ideas are happening on this technology. Technology is a really big part of your career, um, and I always wanted to bridge those gaps.
Speaker 2:Um, because it was kind of like a snobbery around tech for a long time and and you know if because it was just associated with being like quite nerdy in the studio and stuff. Um, now it's like a badge of honor, but back then that wasn't the case. So I spent a lot of time trying to make artists see the other side, um, and that's very much what we're doing at voice. Stuff is to see how it works behind the scenes, from a music industry perspective, but also from a technical perspective, because ultimately it will make you more creative. When you understand how things work, you can start to break them, which is what most artists want to do.
Speaker 1:Well, yeah, yeah, that's it. You captured my imagination in thinking about so many other things, things like legacy artists that are no longer here to represent themselves you alluded to that earlier, you know just being able to kind of um, you know, have them recreated. We saw it with, uh, coachella, with the holograms a few years ago, right, which kind of which kind of then went quiet and then kind of came back again, and with the ABBA thing and all that, because, but it's really about, instead of that being an exhibit for people to enjoy, which is marvelous, it's really about putting the works of those from the past into the fingertips of others on a very respectable and remuneratable way, if that's even a word. You know, could this extend beyond just voices into the musicianship and the patterns and the playing of those that have come before us?
Speaker 2:yeah, absolutely. There's a couple of startups doing that already where, you know, before I joined voice up, the concept that really took my imagination was if I was in the studio and, and you know I wasn't a great drummer, but I would just drum something out, um, anyway, just like the best my ability. And then I had an ai model of, you know, john bonham, and I took john bonham like kind of like a traditional preset, but like the next level of that. Not just the sound of that drum kit, like the preset would offer, but the groove and the feel and the subtleties um, of that playing style were modeled by um, an ai model, with the permission of his estate, um, and I was thinking, god, then you could put it down to, like you know, prince's guitar style, or then you could even go to the next level of that and make it about mix engineering, um, and there's a company called master channel who've done this already. Um, with wes clark, who's like a, a great, award-winning master and engineer in the uk, um, and they've cloned his mastering style and now you can go and use ai mastering to sound like as if wes had mastered it and then that's monetized by him. I do think that's going to become a big, big aspect of um revenue generating for musicians as well, where you can clone your style and all the subtleties that go into that. And the interesting thing is about that as well is while there are, like you know, if you think about wage plugins, there are so many collaborations with famous engineers in those plugins and they create presets, and a lot of that happens a lot with lots of different other plugin companies as well. But those presets are not self-aware, and what I mean by that is I could run a vocal through, you know, a drum preset, um, and while that might sound cool, the preset wasn't designed to go through that sound, and that's just me being creative.
Speaker 2:But with, what ai allows you to do is to apply a preset in context. Because, like this, this is something that I was talking to a friend of mine about a couple days ago. These are the kind of nerdy conversations I have, johnny um, we were saying that if you had a preset and you know, say, for example, um, this engineer, philips at r? Um, he always like, likes to cut the low end by like 2 db or boost you know 10k by 3 db, and that will only sound like him if the input audio is correct in the first place, because if you go in and you go, you're peaking by 6 db. Uh, you know it's not going to sound like philips adar, because the actual input is all over the place and which is why, a lot of the time, mastering bus presets don't really work very well because they're out of context of the mix.
Speaker 2:But now with ai, the ai can listen in inverted commas to the mix and then make a decision about OK, what would Philip Zadar do in this instance to this audio, rather than it being just like just slap this preset. That's static and it's the same for every piece of audio. So you can think I keep saying inverted commas because it's really important not to use language like think and know and learn about AI, because it's not alive, which again could get into, but we won't, um, but it's. It allows people to apply the techniques of a mix engineer onto their music as if they're listening to the music with them and making decisions based and I go okay, what you've actually done there is this is too large, we need to bring that down before I even begin to EQ it or whatever.
Speaker 2:So those, and then you can monetize that. That's a new form of IP as well. We see the voices, ai voices a new form of IP. I do think that musicians being able to clone their guitar style, the drumming style, the bass playing style, the key style, that's a new form of ip that can be monetized as well, and the same with mixing and mastering engineering. So if you're you know, if you're just, you're a really amazing bedroom producer, someone stumbles across your music on soundcloud and they could be like wow, like I really love the way you mix, and you could just sell your ai voice or, sorry, your ai mix um model to them for price of your choosing. Like that's a really interesting idea as well and another way to collaborate with people.
Speaker 1:Yeah, it really really is. I suppose it could be kind of broken down into two different subjects, one being the excuse me, the sonic values and the other being the mathematics of their playability, the probability of their playability, which, you know, is one's composition, I guess, and one's recording, but it's all mappable. I mean, over the years, declan, the amount of records I've produced, I've sat there and thought what would Quincy Jones do?
Speaker 2:here.
Speaker 2:Exactly. Imagine you had Quincy. The thing is it's impossible I hate using the word impossible when it comes to AI, because things I thought would have been impossible are occurring in front of our eyes. But I feel like modeling someone's taste. It's not just about the technique, it's about the decision-making based on the audio that they're listening to. Then they apply a certain technique based on their taste, and that's going to be the real technical challenge. It isn't necessarily like the preset mindset where it's like okay, this is, you know, a bit muddy around this frequency range.
Speaker 2:This Quincy Jones would always cut it here. Yes, that's fine. But what he might say is actually like you need to re-record that because it doesn't have the right groove. Or he might say I've just got this amazing idea for like a couple of harmonies for this part. You know, you're only really ever going to get a very basic version of that because they're, you know unless, yeah, I mean, I don't even know how that would work.
Speaker 2:I think it would be extremely complicated and I know someone who's trying to do this, and they have told me it is extremely complicated to be able to like clone a vibe, if you will, clone an atmosphere in the studio, because when humans work together, they feed off each other's excitement and energy. So how do you clone that? You know how do you clone someone's a bit hungover? How do you clone someone just got off the plane, hasn't slept? You know, someone's a bit grumpy, someone's a bit hungry? All of those things go into the creative process, um, and so you know how do you clone? And all of that is very human and that is why you know, I, I believe that you know, humans will not be replaced, of course.
Speaker 2:I believe that, um, because we have indescribable things that cannot be modeled because we can't even explain toable things that cannot be modelled because we can't even explain them ourselves the creative flow, the flow state. You know what is that? Why does it happen? Why does it happen better in certain circumstances? Why are certain rooms and certain studios so coveted? Why do people feel more creative in those spaces? Or is it just a placebo effect?
Speaker 2:You know all of these things that happen completely unconsciously and instantly when you go into recording space or when you work with your favorite co-producer or your best mate, who you know, just there's an energy between you and that, if you can't describe it, you can't clone it, because you don't know what you're cloning. Yeah, so for that reason, I do think that humans will always be a part of the picture. Of course, if I thought voiceover was a threat to singers, I wouldn't be doing it. No, no, I know what you mean, but I think there's that kind of like human connection that happens either in the studio or with a songwriter. That can't be cloned.
Speaker 1:And coexistence as well. Right, you know the coexistence of AI with humans rather than you know, it's not a battle between the two, it's a relationship between the two. Deton, you're a top guy, I'm mindful of time and there were so many things I wanted to ask you that you kind of just answered already with, like just your amazing sort of ruthless efficiency and thorough rigor so the first time I've been called efficient, I think, in my life.
Speaker 2:So really, oh well, I'll say it again efficient.
Speaker 1:There you go. It's twice now, um, but uh, definitely, I thank you for your time. I thank you for everything that you're doing for everyone and and good luck with all of the, the projects, the ideas, everything that you do and, um, we need you mate, we need you right, so keep doing what you're doing.
Speaker 2:Yeah, we're doing our best mate, but we're not. You know, we're not the only ones. There's a lot of great people out there doing a lot of great work and we're driven by them and hopefully they're driven by us. And, yeah, if anyone you know has any comments or anything to say about you, about what they've heard on this podcast, do reach out to me. You can just Google Declan McGlynn, hit me up on LinkedIn or just email me and we can chat Because, like I said, a lot of these things haven't been decided yet. This is all a work in progress. Feedback from artists, feedback from the industry is really important to make sure that we're going in the right direction, because, instinctually, we feel like we are, but we can always listen to feedback and we can always.
Speaker 1:We need to work, everyone needs to work together on this to make sure it's a safe future for artists. Oh yeah, absolutely right. Yeah, yeah, top man Declan, thank you so so, so, very much uh thank you yeah, enjoy the rest of your day, you too.
Speaker 1:Thank you. Well, there you have it, declan. Wow, what a dude. Um, he's so insightful, isn't he? You know? I should add this I first met Declan only a few months ago.
Speaker 1:I was hosting the South by Southwest London Roadshow uh in in Birmingham and he was uh on one of the panels that I was moderating. And you know, when you just meet somebody, you go wow, they're impressive. You know, I thought it'd be great to have him as a guest on the podcast. Thankfully, we've made it happen and it's just interesting listening to him. You know, it's like he's five years ahead of a lot of people in terms of how he thinks and how he operates and the kind of things, the kind of angles that he's looking at from you know.
Speaker 1:But anyway, I hope that you enjoyed listening to Declan and learning more about VoiceSwap and also, you know, one of the things that I took from this conversation in general and I hope you did too is that, you know, this wasn't just about VoiceSwap. This is about an entire sector and an insight as to where it might be headed or probably is going to be headed in some way or another. Um and uh, you know, and Declan's a huge part of that. So, um, yeah, I hope you've enjoyed it. Uh, wherever you are, enjoy your day, have a good one ahead and until next time, may the force be with you. The Music Business Party. The Music Business Party.
