The Truth About AI Mixing, Mastering, and the Future of Music Tech

Jakob Wredstrøm

1, okay. Let's hope it continues.

David

Seems all right.

Jakob Wredstrøm

Cool. Well, then, yeah. So, trying for the third time. When the episode is done, stick around for a few minutes, just so we can sort of end the conversation and everything can upload if it was a bit slow. The length of the episodes is typically 35 to 55 minutes. That's not really right or wrong. It just depends on the conversation. And I tend to really like...

David

Yeah.

Jakob Wredstrøm

When we talk about hardships as much as possible. The reason why I say that is typically it's very easy to sort of talk about, you know, things we're really passionate about and we'll always do that. So I'll just try to preempt it with like, it is very good for the conversation. People respond really well to if there's honesty, hardship, struggles, including the conversation. Obviously for the reasons that, you know, a lot of the focus on entrepreneurship in music comes from, you know, PR announcements. Uh, and you know, that's really the actual truth. So we try to sort of dissect that here in the podcast as much as we can. Um, and then very important comment that maybe you can keep in the back of your head while we talk is dialogue based conversations are to be preferred. Um, and that's how we have the best interviews. So if I ask a question that maybe could merit a long answer, try to break that up into small pieces and let me sort of do follow up questions. Um, it just. Leaves a better conversation and everything really gets better from that. So that's sort of, if there's one thing you can think about, that would be that thing. Then the rest, I've done a few episodes by now, so I hope I've got you through the process and you can just be you and don't overthink it. Yeah, I think that's it. The last comment is that we normally have like a very simple start to the episode and that is I do a super brief introduction, like five second introduction.

David

Perfect.

Jakob Wredstrøm

And then I ask you, who are you and what do you do? And I asked this to everyone and then you give a 30 second to one minute response, you personally, and then I'm going to ask about RoEx and then you can explain what RoEx does and what it is in like one, two, three minutes. And then we'll basically take it as a journey throughout your career. So we'll go through it like somewhat chronologically, and then I'll try to sort of have these paths outside of the story where we go down into concepts and research. And that kind of stuff. We'll use your Founder Journey timeline as sort of the base of the conversation. And then I'll do like deep dives while we go through that. I might be jumping back and forth a few times, but I'll make sure that's natural.

Jakob Wredstrøm

Great, that was it for me. Do you have any questions?

David

No, this is go for it.

Jakob Wredstrøm

Great. Then I will start in five seconds.

Jakob Wredstrøm

Hey guys and welcome back to the Sound Connections Podcast. I've been looking forward to this episode. I hope we can get nerdy. David, welcome.

David

Thank you, thanks for having me, how you doing?

Jakob Wredstrøm

Yeah, I'm preempting this conversation. Let's see how an entry gets. But David, you work with something really interesting. You have a company called Roix. I've heard your company name thrown around in so many different constellations and conversations. So I've been looking forward to this. But David, for the people who don't know who you are, who are you and what do you do?

David

Yeah, sure. Yeah, I'm Dave Ronan. I'm the founder and CEO of Roex. I've worked in music tech well over a decade now. I'm from Limerick City, Ireland originally. Yeah, I've worked at Native Instruments. I worked as AI Music and for the last three years I've been building Roex. So yeah, lovely to meet you.

Jakob Wredstrøm

Great. And also at prestigious positions like head of research from AI music and that obviously sort of came to an acquisition of the company. We have a real expert in the studio. So I'm looking so much forward to this. But could you explain more detail what is Roex? What do you guys do?

David

You

David

Yeah, at very high level, what we do is we build tools for musicians and composers, content creators essentially to give them the confidence they need to release their music. More specifically, we're essentially democratizing the audio production process. But more specifically, that's AI driven tools for multi-track mixing. We're the first company to bring any kind of multi-track mixing tools to market. We also do mastering, we do audio QA. We do audio cleanup. We do audio restoration. Any of the kind of technical challenges that come with audio production, we have like a Swiss army knife of like essentially tools for dealing with it.

Jakob Wredstrøm

And how do users use this? It primarily enterprise software? How do this play out?

David

Yeah, good question. So we have two kind of arms to the bit to the business. We have a B2C or D2C business and then we also have a B2B business. So in turn on the B2C side, we have our flagship platform, which is called AutoMix. Essentially, a musician can come along, upload the recordings, provide some very basic mix settings and get a mixed and master track back in minutes instead of days at very low cost. We've also integrated into some of the major DAWs out there like Ableton Live, Bitwig. Presonus Studio One, actually we've got Cubase coming as soon as well. That's our kind of main platform, but then we also have another one called MixCheck Studio, which started as a free kind of tool essentially. You can come along, upload your MixMaster track, we'll analyze it, tell you what's wrong with it and how to fix it in your DAW for free. And we also have this additional tool that we just launched before Christmas that we'll actually, we know what's wrong with it, we'll actually fix it for you in most cases. But then obviously we have a B2B arm as well, that's basically all those tools essentially wrapped into APIs and an SDK.

Jakob Wredstrøm

Mm.

Jakob Wredstrøm

Amazing. Well, we'll dive into that and dissect the purpose of that throughout the conversation. But David, you must have come from music, I would assume. So can we just hear about how do you get into music and get this particular interest?

David

Yes.

David

Yeah, sure. Mean, it all kind of goes back to when I was about 15 or so, I would say. Really big into electronic music. That's kind of my bread and butter. I mean, I listen to like lots of different kinds of music, obviously like hip hop and prog rock, you name it. But like I'd say my main thing is electronic music. I was always a big fan of stuff from Warp Records growing up, like Reflex, IDM. Techno, jungle, like big affix, twin fan boards at Canada, square pusher, all tech, that kind of stuff. Yeah, I would say that's pretty much my bread and butter. Guess experiments, very experimental kind of electronic music. And yeah, I guess it all started when I first opened up the DAW, which in my case, it was Fruity Loops, which is, I think it's FPL Studio now. Yeah, have you, you're familiar with it? You used it before? Yeah.

Jakob Wredstrøm

Amazing.

Jakob Wredstrøm

Yeah, absolutely.

David

And I remember basically when I opened it, was just like, know, what the hell do I do? How do I, how do I, what's MIDI? You know, why do I have to use a compressor? Was, you know, it's just, extremely overwhelming. But I come from a very technical background, I guess. Like I studied mathematics and computer science. And I suppose, I quickly found that like I was probably better at the technical stuff when it comes to music production and the actual making music itself. But I had loads of friends who were very talented bass players and guitarists and whatnot who were not very technical. And I guess I was able to help them, you know, get over those challenges because I come from a very technical background, essentially. So that kind of led me to studying a master's in music technology at the University of Limerick. So I'm actually from Limerick City originally, which was a really fun year of like just applying my technical background to really technical aspects of music and audio. And I guess that's where I learned how to mix and master. There was obviously a module in that kind of things. And I guess it was around then that got me thinking, hold on a sec, there's a creative side to mixing, there's lot of this parts of it that aren't necessarily creative, they're more mechanical things you have to do. And that got me...

Jakob Wredstrøm

Thank

David

Thinking then, because mean AI hadn't become, you know, center-focused then, it was like very kind of early stages of that. So that would have been like 2010, I think. So yeah, it's gone back a bit, but like that's kind of where I suppose everything kind of started, I would say.

Jakob Wredstrøm

Yeah, and it's interesting and you know, I have a lot of opinions about music production, audio engineering, that's sort of my academic background and the first part of my career was also sort of my life. And I think it's very interesting and I'll just make a commentary about sort of AI in general. Now I'm not talking. Necessarily what you work with, but AI in general and how it's disrupting the perception or process of music production and all the subsequent steps or steps before that when releases come out. And things are changing. Like my own story is I start with music. Fairly late, I was 17, I didn't really have a relationship with music and I fell in love with producing and mixing for some reason, it was a weird angle. But I produced a lot of music and ultimately I stopped when I was 25 because I couldn't make enough money because I was very, very good at my craft but ultimately I needed to pay musicians to come and produce music so I abandoned it just because this is not going to be my livelihood because I, you know, I might be really good at executive producing, like this traditional kind of producer, but that's not financially viable. Ended up being an entrepreneur, I'm happy with that. The problem is that the barrier of entry for music production, audio processing, creating music is so incredibly high, literally still to this day.

David

Yes.

Jakob Wredstrøm

And which means that you need to go through these traditional years of craftsmanship in order to get like a even like a close to like okay commercial product. Like I know people who have been doing it professionally for 10 years and I still look at their here the stuff I'm like, yeah, ish maybe, I don't know. It's it is it is that hard, you know, as a craft and

David

Yeah, totally.

Jakob Wredstrøm

So there are some things that definitely can be addressed, I guess, in the accessibility. What's your philosophy behind that? You know, when it comes to the barrier of entry, what you're working with, what do you want to achieve?

David

Yeah, I agree totally with everything you've said. I mean, the barriers are extremely high. Mean, it's a very, especially if you're like come from a very creative background, you're completely non-technical. And then suddenly you have to do mixing and mastering. You're like, as I said, like what's a compressor? How does EQ work? You know, these are all very technical things, balancing frequencies, dynamics, you know? Mean, a compressor was mystic to me for quite a while, you know, until I really kind of dug deep. And that's me and I come from a technical background. I can't imagine what it's like for someone who doesn't come from that. They're just like, what the hell? Especially the thing is, now especially, you've spent, like you could say, you've met people who have worked on this for 10 years and they still don't sound really great. Imagine you've done 10 years of being an amazing lead guitarist and then you want to get your music out there and suddenly you're like, what? I have to do another? Couple of years learning how to mix and master for this to sound good, know, it's not really sustainable. So I guess in a way that's kind of like what we've been doing is to kind of lower those barriers to make mixing easy, to be able for you to get to a rough mix or, know, I mean, don't like, think what like, yeah, to get you to a rough mix as quick as possible, basically. So you can send it to someone, get some feedback and it doesn't sound like absolute, you know, crap essentially. Because yeah, it's just really high barriers of entry for sure.

Jakob Wredstrøm

Mm.

Jakob Wredstrøm

And what's going be the ripple effects of, know, now I'm sort of going a bit to the conclusion of the company before we get really into the conversation, but I think it's interesting. What's the ripple effects you see if we really provide the tools of automixing, automasting, at least to a level that, you know, you can take something that isn't to something that's quite good. Like what's going to be the positive effects from that?

David

Well, I guess it's more, yeah, mean, it's like suddenly, I mean, think of it this way, like if you wanted to get a professional sounding track mixed and mastered, I mean, I know, like what we typically do is you go on, you go into a market, not unless you know someone, but like if you go on, you go onto Fiverr and like it's about $120 to get a track mixed and mastered. And suddenly now you can go into our platform and get it done for $5. So suddenly, you know, you're making it easier and more accessible to people who don't necessarily have the means to get their stuff mixed and mastered. You know, that's one thing. And I guess the end game as well is like, how we kind of, actually, the other thing is, we kind of started the company was we were initially, it was kind of like, give you back a mix and that was it, and you take it or leave it. We kind of quickly found that's not necessarily the way to go about it. I mean, it was kind of obvious, like, obviously you need to test stuff as you go along. But we found that like, we don't really want to be a replacement for the mix engineer. We want it to be assistive. So essentially now the tool is very much an assistive tool. It'll get you to where you need to be. It'll do all the of the mechanical stuff. It'll bring you to the 90 % point and it leaves that 10 % there for you to be creative. Essentially the whole point of the tool is give you more time to be creative because that's the fun stuff. And it removes all the mechanical boring stuff essentially. I think that's kind of been the kind of the ethos of what we do essentially.

Jakob Wredstrøm

Mm.

Jakob Wredstrøm

I might be asking a bit hard, maybe a bit cheeky question, but I'm curious about this. So let's just assume that achieving the perfect mix using AI was possible. Why? Well, let me frame the question a bit differently.

David

Mm-hmm.

Jakob Wredstrøm

Why is your purpose not to achieve the perfect mix? Is it because of the technical limitations? Is it because of sort of industry impact? Is it the business model that wouldn't hold up? Like why are you talking about assisting versus a complete exchange?

David

Yeah, agree. Good question. So I mean, the thing is mixing is so subjective. And the other thing is like, ultimately music is about conveying emotion and being able to like sit down with a mix engineer and go, look, this song is about my ex-girlfriend. And, you know, essentially this chorus is like, this really needs to, you know, hit basically. How are you going to explain that to a machine? You know what I mean? Like, you know, it doesn't really translate. So I think

Jakob Wredstrøm

Mm.

David

If a mixer can suddenly be like, look, can just ditch it, get the mix to like the 90 % point. And then they're talking to that artist and go, they're just doing that extra 10%. That's a win. You what mean? That's essentially, you're retaining the human element. You're returning the artistry. The emotion is still there. I think that's the more desirable outcome, know, essentially keeping the human in the loop because music is a human thing. You know what I mean?

Jakob Wredstrøm

So to some degree, how I hear your answer is that your assumption is that technology will not be able to or is very far from changing the place of the interpretation of the human and putting that into the mix. Questions to that, you know, I just saw a new video model. yesterday in China and one of the big promises from their side is prompt adherence. What you convey in prompt adherence, so what you convey in your prompt, in your writing is the output stays very true to that. So if we talk about the concept of prompt adherence when it comes to audio mixing, again just assuming hypothetically everything is possible, don't you think that could be translated through AI?

David

Is what, sir?

David

Right.

David

Yeah.

David

Maybe for a very simple mix, but I just think there's so many elements and so many things kind of interacting and moving within music essentially. know what mean? It's not just, it's not just like the, how things sound good and how polished they sound. It's like, you know, where the melody sits, when the melody plays, that kind of thing, or how they kind of sit in the mix as well, think is important. I mean, theoretically, yes. I mean, if you take the limit of whatever it, out to it goes, I mean, I'm sure. It could at some point, but I don't know for anywhere near it at this point. But I could be wrong. know, who knows what's gonna come out in the next five years. The way things are moving so fast right now, I think it's entirely possible. That's the other thing, you know?

Jakob Wredstrøm

Hmm. We'll go back to it started a minute, but again, I said I was looking forward to the nerdy talk. So I'll just continue a bit on that, on that angle. we had a really interesting, guest here. a few, I guess a year ago, a professor, in neurology and he was speaking about the perception of music when it came to AI. And we talked about a few points and he said, one of the big things of perception of music is acoustics. And, you know, theoretically.

David

Mm-hmm.

Jakob Wredstrøm

AI can achieve optimal acoustics when it comes to compare to, you know, human produced music. And then there's the sort of the cultural understanding or the association to the cultural understanding, which AI will struggle with if it if it is exclaimed that this is not human. But again, if it is not, you know, let's say again, we're just hypothetically, if everything was mixed perfectly,

David

Mmm.

Jakob Wredstrøm

by AI and people didn't know that, that necessarily wouldn't touch that. So the cultural understanding of who made this, what was the context, what was the process? And then the last thing he talked about is what is the community engagement around it? So, and again, in my, when we talked about music, you talked about obviously fandom and subcultures where that would be where AI would really struggle because that is basically very, very human. But when it comes to mixing,

David

Yes.

Jakob Wredstrøm

Is that a thing? Is there a community aspect for the listener, for the fan, where community and the process of mixing really implicates that? is AI in mixing able to really work in all those three areas and not take away worth of the end user's perspective?

David

Yeah, I it's interesting because I did some research before. well, yeah, actually, so I did a research paper when I was doing my PhD and it was basically about the emotional impact that mixing has on people. And what was interesting about it was that unless you have critical listening skills of say a mix engineer or someone who's worked in kind of audio, you might, like to the, basically what I'm trying to say, actually just step back, it's like. To the 90 % of people who don't have critical listening skills, the emotional impact of mixing might not necessarily have much of an effect, but it does on people who do have those skills. So maybe this kind of stuff would work for the, mean, like having a good mix might not be that apparent to basically the general public, but the people who do really understand listening music will definitely notice the difference essentially. So I think that would be one way to look at it.

Jakob Wredstrøm

Hmm.

Jakob Wredstrøm

It's very interesting. We'll touch more on it. We'll go back to your story because obviously that's also a really big part of this conversation. Okay, so you sort of went into audio engineering, got really curious about that, but where did your career actually start with, you know, carrying out your research or the practicalities of what you working on?

David

I guess when I finished my masters, I did a yeah, it was quite a technical masters and I was just like, always like, and then I moved to London for like broader opportunities. I always knew that I wanted to go back and continue on the music tech path because I kind of drifted off into doing kind of other engineering stuff. And then I'd heard about Queen Mary University of London, which is one of the largest music tech research groups in the world. And I heard about Professor Josh Rice, who's one of the founders of lander and it was all and he was like, was basically he was, he had a research group that was focused on intelligent audio production tools. And I was like, I remember I, because I really enjoyed learning how to mix and master during my masters. I was like, okay, this is perfect. I really, I'd love to have a chat with him. So we had a chat with him. I had a chat with him basically, and he told me about the topics and I was just like, this is, you're solving the thing that I had thought about and I was very interested in my master. So I applied basically, yeah, ended up working at Queen Mary for four years. and doing intelligent audio production tools. And I would say I did a very, it was a technical PhD, but was also, a lot of it was talking to mix engineers, understanding their workflow and how they kind of have essentially how they work because essentially that's what you're trying to replicate. And it was looking into like how emotional mixing can be as well, which I mentioned already. So I got a really nice kind of nicely spread out balance masters, but like it culminated in me building it basically. an AI mixing system by the end of it.

Jakob Wredstrøm

Yeah. Somewhere in between there, I'm sort of a bit in doubt about the chronology. You also were at AI Music and you were at Native Instruments. Where did this fit into that whole picture?

David

Yeah, good question. during, in the middle of my PhD, I took an interruption basically to go work at their research team in Berlin, in their offices in Berlin. So that was there for six months. Had a really good time working there, learned loads. Yeah, I was working with some of the core, real core members of the team who brought some of the great, like a lot of the great products that they brought to the world. Yeah, it was just like, because was getting to like take a lot of the theory of machine learning and all the stuff I'd learned basically and actually applying it to building really, really cool tools. wasn't, AI music didn't happen until I finished my PhD. Basically, I finished my PhD in London and I had obviously very little money because PhD life is quite tough in London. And I had no startup experience other than obviously Native Instruments, which I don't think you would class as startup at this point anymore. But yeah, I was there for four and a half years as their head of research, which was, yeah, it was amazing because I got in right at the start. Even at the start, I think we were doing B2C before we kind of pivoted to doing kind of B2B. So I was there from the very, very start where it was an incredibly small team doing, figuring out lots of things and experimenting and just exploring to like right up to the end where we're like, we're very focused on what we knew and we were building. And yeah, mean, obviously take a step back. What we were building was, sorry, what we were building was systems for generating music, but like we didn't, wasn't like the Gen AI systems that we know today, like Suono and Udio. What we were doing was we were taking, we had a team of music producers who were creating loops and samples. And then we were using AI to essentially stitch them together, still retaining the human element of like music creation. and creating new music out of it. So you could go, okay, give me a 45 second house track in the key of A minor at 130 BPM and it would basically generate it for you. And I would like, okay, want that. mean, you'd be able to tell it where the drop would come and it would just do it. And yeah, it ultimately, it was very successful. We ended up being accorded by Apple, which is pretty awesome.

Jakob Wredstrøm

Amazing. That's great. you sort of, I've tried to study that story and in all honesty, I've been a bit confused about AI music. Like I think I haven't really fully understood, I guess because you've pivoted and I'm trying to do my own research, but it also seems like for me from the outside, and I might be wrong here, that it's almost like a happenstance kind of company where like building something, then you just became really, really good at it. And then the acquisition for me, I just don't see the sort of logical startup journey transition there, which was just you built so good that was interesting, what was the incentive there?

David

Um, yeah, I mean, I guess we were unique in that, think a lot of the other companies at the time in the space, like you had Juke Deck, you had Ampere. Uh, obviously there was another one I can't remember off the top of my head now. They took a very different route to us. think they, like, they were kind of composing, they were going, like using AI to basically compose and generate MIDI. where we didn't go down that route at all. We went very much like, okay, we're gonna take human composed music and make it stitched together, like I said. So that retains the human element of it. It didn't sound like a computer essentially had made it, which I think that was what really, I think stood to us. That was the kind of like, us kind of stand out essentially. Yeah, and I think like how we were promoting it and how we were kind of getting it out there was,

Jakob Wredstrøm

Mmm.

David

we had a YouTube channel which was just broadcasting like this music 24 seven. And I think that's how they say, and that was just, yeah, it's like, look, this is our music and you could listen to it like, and it would just be constantly evolving and changing. Like we'd like a, think we had a lo-fi, of course we had a lo-fi hip hop channel. We had a house channel. I can't remember, was a years back. Yeah, we basically all these different genres and we're just pumping it out 24 seven.

Jakob Wredstrøm

wow.

David

And I think that's how it was discovered basically. And we're like, wow, this sounds really good. And they were obviously able to dip in at any time of the day and the music would be completely different. It would evolve. We were doing, and we were like, because we were able to generate the tracks at any kind of length or any kind of genre or sorry, key or tempo we wanted. We'd also put in these kind of like really crazy DJ effects. So like it was just like a big, insane, long 24-step DJ set essentially. Yeah, and I guess that's kind of how it came to be and how we kind of, think we got discovered essentially.

Jakob Wredstrøm

Getting acquired by Apple, again I'm just making assumptions you can confirm or say if it's not correct. I would understand that as you know they would acquire for technology rather than know revenue per se. there's something about what you guys did that was interesting for them and obviously you know research is a big part of that. So what was the incentive? I don't know if you can comment here that in specifics like why would they be interested in acquiring this?

David

Yeah, I mean, I can only really speak from a high level because obviously a lot of stuff is confidential. like, I mean, it was very much the technology and the team they were very interested in. We were definitely, we were generating revenue, but I think, yeah, for them, I mean, it's kind of Apple's typical playbook, it's tech and team I think they're most interested in. And the vision and where we were going to take it because we were like, because we were like pivoting for a few years and figuring out like it was only kind of

Jakob Wredstrøm

Yeah.

David

when they came along that we were like, okay, this is the way. We had our path essentially, so, and the visions aligned, I would say. That was kind of what worked out well.

Jakob Wredstrøm

There is a big leap from being a researcher, like PhD, job, know, start up obviously with AI music, still as a part of a team and then going on your own to found a company, that's a big step. And I think, know, it's not far removed from academia, but to some sense it definitely is. So why did you do that?

David

Yeah.

David

yeah.

Jakob Wredstrøm

Why incentivize? What was the thing you wanted to do that was different?

David

Yeah, mean, yeah, that's a great question. guess, like, because I was working as an engineer before I went back to the PhD, which kind of deprogrammed me from thinking like an engineer to think like a researcher. And then when I went back to AI music, it kind of reprogrammed me or deprogrammed I don't know where we went with that, to be more of an engineer than I was a researcher, but still having that kind of research background, essentially, and knowing the right questions to ask and really understanding the technology. I guess the thing was, I knew at the end of my PhD that I had this intellectual property around mixing, and no one had done it before. Obviously, AI mastering was done by Lander. There's been many people who've copied it since, but no one had done. So I knew I had something unique, but I knew that I didn't have the ability to start a company then. I wouldn't know what to do. And I think having worked at AI Music was like the... perfect opportunity. mean, the founder of AI Music, he previously exited companies before for very successfully. So being able to learn from someone like that, I think was very attractive for me. And I did learn a lot from the founders, essentially. And basically by the end, I knew, okay, I've worked at a startup that's successfully exited. Now is the best, for me, was like, this is the best time for me to start my own I have. I know what works and what doesn't essentially. So it gave me the confidence to do it. So yeah, I turned down a job at Apple basically to start Roix, which was hugely risky at the time. had a nine month old daughter. mean, my partner was just like, are you sure? I was like, I'm pretty sure. mean, obviously at the start it was extremely daunting. mean, starting a business is tough. mean, no one knows who you are. You're just posted. stuff out LinkedIn hoping it sticks, you don't know what works, you don't understand marketing, you don't understand BD, there's so many things you just don't get and yeah I mean I'm so glad I did it obviously because the company's successful today but like it's scary it is but yeah I don't think I would have done it without having worked at a startup I think that was definitely crucial in basically what I'd done.

Jakob Wredstrøm

That's amazing. And I can recognize myself. I jumped out as a full-time entrepreneur when my kid was, I guess, two months old. So I was equally as reckless as you, my first kid. The difference was that I didn't really have much going for myself. Like I've always been pretty good at what I do, but like, I didn't have a shining career. For me, that was just like, know, the next thing in line. I sort of, I kept being stupid for enough years to something to make sense now, but.

David

Yeah.

Jakob Wredstrøm

Yeah, but like what did your partner think about this? Like for real, like you come for something extremely successful. And yeah, of course you might be sort of, you can argue I've been at this company of Seadeck and it happened and there's sort of this backstory, but the reality of entrepreneurship isn't that until it happens. So there must've been some internal conflict there, guess.

David

Yeah, mean, it was more like the way I saw it and the way, and actually she very, supportive obviously that I did it. But like, it was like, okay, I could go to Apple and I'm sure it would have been a very, it would have been a well-paid job and it would have been a very cushy existence, I'm sure. But like I knew, and she agreed that like after I'd be kicking myself after four years that I didn't, that I know that I would live with the fact that I never tried. I think that was it for me. And the way I saw it as well, was like, look, what's the worst? One of the worst things that happened is like, you try it, you fail, but at least you tried and you have X amount of years of the experience of trying and running a business. And at least those, will have, like there's skills and things you learn there and life experience that no one can take away from you. I think that was my kind of mindset going into it. I think I was lucky as well to have obviously a very supportive partner. And having a supportive family and very supportive friends that are just like, know, just go for it, you know, essentially, you know, if you're gonna, like I knew I'd regret it if I didn't do it. And I'm so glad I did. And I think you probably felt the same way.

Jakob Wredstrøm

Yeah, I can also imagine like you personally, you know, being surrounded by a professor founding Lander or founding partner of Lander in Native Instruments and AI music, like you were exposed to start-up-ing enough that you sort of also had a sense of what you went into, I guess. But how does this sort of, sorry, you can respond to that.

David

I just going to say, yeah, I mean, I will say one thing and I think it's important. I was lucky, like not all people have the, like the, I had the lucky thing where, not the lucky thing, but I basically, I had some intellectual property for my PhD. So like, wasn't going into this with a blank slate. I had an advantage. That's the other thing. Like, I'm not going to say that like, was, like a lot of people, they're just building stuff from nothing. That's, that, that could add on an extra two years. You know, I had something so that, that I did have an advantage. So that's also worth saying. I'm not, I'm lucky in that respect.

Jakob Wredstrøm

Yeah.

Jakob Wredstrøm

Yeah, no, that makes sense. That makes sense a lot. But how did it progress? Okay, you sort of go out of AI music and then you start this company. Did you, how did you sell? Like, what did you sell? How did you build it?

David

Yeah, mean, yeah, because like the, yeah, was like, basically, because the stuff in my, was like, like obviously the stuff I did in my PhD was like stuff that would run in a lab. But like some of the IP was valuable, a lot of it was wasn't basically. But like it was stuff that like you had to run on a computing cluster, which could give you, like I think this, like the original prototype could only work in mono. It could mix only 20 seconds of audio, could only mix like four tracks at once. It was absolutely useless to anyone. So that was the first, yeah, absolutely useless. So I was like, I would take a day basically to give you results. So the first challenge was like to get that into some kind of usable format. So that was basically taking all that lab code. And I obviously had all these ideas of how I could make it into useful product and converting it all into C++. Like I've been programming a long, time. I think I started programming when I was like nine or something like that. So I'm fairly adept at that. But like it's still a challenge. Like we didn't have chat GPT then. So was like having to like dust off the C++ skills and basically just builds the essentially like an SDK or it was kind of like half Python half C++ structure to basically turn it into an API. That was my first, my first like task was like, okay, this needs to be in an API. I knew absolutely nothing about web programming. I'd never built a bloody web app in my life or anything like that. So I like. Obviously came from a very R &D background. I knew how to do DSP, I how to do machine learning, I knew how to build programs, but to do anything like customer facing like that was like, forget about it. So yeah, was like having to learn how to become a web programmer was the first thing. Yeah, that was the first thing, learning how to use Python, Flask to build the app. Then it was like, okay, do I have time to build React, to learn React? No, I kind of... Because like that was the other thing I was reading so much stuff online about startups basically just like it's the quicker you can get to market the better. Because of the feedback loop because people will tell you it's crap or what to improve. So I was like, okay. Do I learn react? No, okay. I'll just do it in HTML CSS. And you know, and that was that was a like I basically built this like extremely basic Web app. That like had no like snappy nice web apps that were used to now and just launched it.

David

And met some very great people along the way who helped me kind of like shape it in terms of like, in terms of a product like Ed Newton Rex, the founder of Jukebox, Juke Deck, he helped me a fair bit at the start, gave me some really wise kind of words and what to do. But yeah, it basically launched it. Somehow I just did a lot of guerrilla marketing on Reddit and stuff like that and managed to get my first customer on the B2C thing. That was such a great feeling. And it was like, my God, someone's actually paid for it. You know, someone's actually made a mix, they're happy with it, they paid for it. And that was what allowed me to do a pre-seed round. Yeah, that was it. That would have like all kicked off in January 2022, sorry, 2022, yeah. And I think I didn't get like first customer till September or something like that year. And then we did the pre-seed round at the end of that year. So that was kind of the thing. And then had to put the, also the other problem was, was like I had to spin out from the university because some of the IP belongs to university. So I had to like get all that stuff squared off with the legals and stuff like that. That was also a very challenging and tedious part of the journey.

Jakob Wredstrøm

How did you finance this? Is this from savings? Because I can imagine one thing is you working, know, compute cost and like all of this doesn't sound cheap. Like how did you do that?

David

Yeah, partly my own cash. I put a lot of my own cash up front, which I was then later to kind of be able to recover some of it from the business. The university was able to give me a grant of 50K. I think can't remember what that kicked in a bit late, but I was still able to kind of recover some costs from that. But then it was like free credits from Google definitely helped a lot. Yeah, I mean, just, yeah, was mainly like the first year it was, yeah, Dave eating ramen noodles, you know.

Jakob Wredstrøm

Yeah, okay.

David

Not actually but generally it was like low cost living. The only out goings we really had was nursery fees for our first daughter. The first year was tough but we made it work. Grants and then our pre-seed funding.

Jakob Wredstrøm

And it is a Did pre-sale funding come from industry people or who understood this potential?

David

Yeah, was at a UK VC called Hatch Ventures who've been absolutely amazing. They reached out to me. I wasn't actually really raising at the time, but I was like, look, they're out and had a meeting, it went really well. They really liked what we were doing. I mean, the other thing is like I would say they really understood the potential that like we were a spin out. The tech was already there. It just needed to be commercialized. So I think that obviously really helped as well. Yeah. Yeah, they introduced me to some other people in the music industry who gave me their blessings, which really helped as well. And they invested in the university then because the university was like, okay, there's actually institutional money here will follow on. So they put in money as well and it kind of just took off from there. And they've been really great, fantastic investors. Anyone in the UK, would really recommend reaching out to them because they're good bunch.

Jakob Wredstrøm

One of the big questions that I'm sort of left with of this whole era of AI is the implication of the creators. And I've had many discussions about this on the podcast about sort of what is the creator application when it comes to finances, the potential opportunities in the future for craftsmanship. And that has been very interesting and also, you know, quite, you know, one-sided to some degree conversations when it comes to like, you know, you know, an audio like what's that gonna happen? But this case is a bit particular, right? Because you basically, it's a middleman function you're touching, and it's effectivization, and it's sort of making something from the rough to be a bit better as of right now. But what's your outlook on the consequences of mixing engineers on this type of technology being out there?

David

I think that it will not be good for the people on Fiverr who are charging $120 per mix but they're only just basically taking your multitracks and using some Ozone plugins and firing it back to you. I it's gonna be good for them because, and that's fair because they're just, you they're not actually doing much craft, you know, they're just basically running through Ozone and whatnot. But I don't think it's going to affect people who are good at mixing and that's their craft because... Think about it this way, and I had an artist tell me this I thought it was beautiful because it makes so much sense. It's like if someone goes, okay, here's two grand, this is for you to make your album, or 10 grand, or whatever it is. It's art at the end of the day. You're not, okay, I've got 10 grand. I'm gonna go hire the best and the best mixer, the best mastering engineer I can with this to make it sound as good as possible because my art is important to me. But if I can afford it, in the case of most people, you can use something like our service. Know what mean? I don't think like, essentially like it's, our service is there for people who can't afford it, but it's also like if you can afford it and you do have the means to do it, you are going to just hire the best engineer you can. You know what I mean? You're gonna work with the rich cutters or whoever in the world. You know what I mean?

Jakob Wredstrøm

No, I agree. And that's sort of the general sentiment we've had on the podcast. Sort of me playing devil's advocate on that is that really the biggest issue in the creative economy as of right now, which will also be even more so with AI in general, is that the midsection. Again, know, I early in this interview referenced like 10 years of experience and still feeling like people who are, you know, maybe blasted, master's level. Students in music technology that's trying to get a career from mixing. They might previously have had some opportunities of like, okay, pay me 150 pounds and I'll do your mix. And they can make like a scramble living out of that can motivate, you know, a further career progression over time. AI comes and disrupts that a lot. So what's your thought about that specific midsection?

David

Yeah.

David

Yeah, but I think like, I know where you're going with this, but I feel like these tools are there to make you go faster instead of actually doing the job for you. Know, suddenly if you're able to get to the 90 % point a lot quicker and then you're doing the 10 % of like just making sure it all sounds perfect for the artist, you know, right, normally you can do, I don't know, 10 mixes per week. Suddenly you're able to do 30 mixes per week. You know, you can, and that's, you're saving yourself time. That's more revenue essentially if you're, that's your, freelancer and that's how you do it.

Jakob Wredstrøm

Mm.

Jakob Wredstrøm

Hmm.

Jakob Wredstrøm

Hmm.

David

So I don't think, I mean, it's like any tool, you know what mean? Any of these tools, like when Ozone or any of those can be like, they made you go faster essentially, you know? Like automation, like in the door, the door came along, you know? Suddenly you're able to like automate all these things. It's, you know, I think it, yeah, I mean, this is the thing. Don't think, like some, maybe some aspects of the job will go away. Maybe that's a good thing, but like, I don't think it's gonna take away from the creative aspect of it, you know?

Jakob Wredstrøm

Hmm.

Jakob Wredstrøm

Yeah, I think to piggyback that statement, I think it could level the playing field. Let me explain as to why. I had a conversation with a, not in the podcast, but in a business context, a well-renowned mastering engineer, I'm not going to mention by name, but we were talking about their business, the business model and sort of how they break it down. And what that person explained was that in reality, most of the masters to do is a five to 10 minute thing. It's they have a setup that works really well. They usually speaking have clients as really great mixing engineers. And what people pay them for is their very specific angle, maybe their setup. That oftentimes is like a chain of presets that sort of works. And then an opinion that matters, and then that delivers a good result. But in reality, most of the great, both mixing engineers and mastering engineers gets quality that is already incredibly good. And their effort is very much the autistic touch, or the quality stamp, or the security. There's many elements there. But most of the midsection, it's really about making

David

Mm-hmm.

Jakob Wredstrøm

Something that's okay to be good. And there goes a lot of effort into that. So to some degree, agreeing with your statement, it could also just be the talk of just loving the playing field, because this is how the pros work anyways. They're not fixing stuff. They're enhancing what is already good.

David

Yeah.

David

Yeah, even so, like it definitely levels the playing field for sure. But it does allow more room for creativity and expression. You know what mean? It'll change how people mix. Imagine they're spending more time actually doing the creative aspects of it than, you know, they're suddenly like, okay, can actually do more interesting things here. I'm not actually having to deal with bloody setting all the kind of basic EQs and stuff like that, you know? So it does, yeah, I think that will probably be the, you know, culmination of it.

Jakob Wredstrøm

Mm.

Jakob Wredstrøm

Think that's really interesting because again, that is where the bulk of time goes usually. Is the solving all the technical stuff. You can, you know, I've done so much mixing and I've worked with so many mix engineers. I've even been manager for two mix engineers. So I know the of the economics around it and the amount of time they spent on just technical stuff and then like maybe a fourth of the time they can use in creative. It is ineffective. At the end of the day, what is interesting is living out the vision of the answers and technicalities are oftentimes things you just need to get rid of. Like you need it to be balanced. Need it to, you know, not sort of overlapping and, you know, frequencies being muddied or you don't have space or dimensions, like all these things, even though it sounds like it's a creative thing, really is it, it is a technical thing. And there's so many things you can solve technically that will express itself in a stronger creative expression, but still the actual creativity happens in, okay, when things are taken care of, now I can think, now I can draw some lines. Yeah.

David

Yeah.

David

You can play around.

Jakob Wredstrøm

So I agree to that statement and for me it's also comforting hearing it from you because obviously I wouldn't say I'm worried about the future for creators but I'm very curious to see what it does and it really happens my reaction is very much of sympathy to some degree because I studied at some of the best conservators in Europe and

David

Mmm.

Jakob Wredstrøm

I've seen my colleagues or whatever you want to call them, fellow students, know, working with the craft of 15 years and the ones that does not have the adaptable mindset of what creativity will be in the future will be victims of technology at large. And that's my worry. But again, that is also the nature of any technology in any industry throughout time. This has happened over and over again. Yeah.

David

Yeah.

David

I mean, just an interesting anecdote. Was chatting to a colleague of mine yesterday about kind of obviously where AI is going and he was talking about Sunu and Udio and we don't have, I'm sure it's been talked about. We don't need to go into like the ethical side of it, whatever. Like what he's seeing and what he's hearing a lot, cause he's a mix engineer and he's seeing a lot of artists that he worked with, they're using Sunu and Udio. Generate something, but they're not taking the full track because they're artists themselves and they're creators. Why would they? It's just a gimmick. Essentially it's self, it's quick, you know, it's instant gratification. What they are doing is getting the stems or the source separated stems and they're essentially sampling parts of it and using that within their works. I think that's interesting because essentially it's just sampling when you think about it. I mean, obviously we don't have to talk like, I mean, let's just assume these are all ethical and we're all good, but like. That in a sense is the same thing that we would have done before. You would sample hip hop records and old funk and soul for drum and bass. And I think what's interesting is like for me personally, as someone creating music, I'm not very good at programming drums. But being able to have tools, can quickly, I can just type in, prompt something. I need to kind of like boards of Canada, lo-fi, blah, whatever. And I can sample that. Essentially I'm not then having to troll through loads of like Original recordings and I look for things essentially the same thing. I think that's really cool. I think that's interesting. And how things are going and it's making life. It's just making life easier. You're suddenly able to get to where you want to be. Faster essentially, you know what mean? But obviously there is the fact that sooner or you haven't been trained on all the good stuff. But we don't need to have that discussion obviously because it's been discussed to death.

Jakob Wredstrøm

I definitely.

Jakob Wredstrøm

Yeah, it has been discussed to death. Obviously there's new changes in the story now with licensing, equity and ownership and lawsuits, but again, that's a different talk. I think this has been incredibly interesting and I'm very happy that it has been nerdy. But before sort of we close up the conversation, David, I would just love to know like, where are you bringing this company? Where do you see the future hold for the technology, your company, what's gonna happen?

David

Yeah, mean, guess like we, yeah, mean, we're, the moment we've just like, we've just launched an SDK. So we're, we hope to be proliferating through as many DAWs as possible in the future. Cause it's to be used on mobile devices, but more in a kind of a actual use case. I like we're very much been focused on music. We love music that we come from music, but like our tech has immediate applications in film and TV post-production, not necessarily. Even film and TV like for UGC creators, because obviously there's audio that goes into that. So the tech is very applicable there. The same problem, a lot of the same problems exist, but maybe it's more dialogue focused. So we're planning to take it into that next. And then obviously video games as well, because it's the same problems. Where it kind of gets these were like, mean, the other thing we've kind of bought it up against a lot is like. And you probably hear this yourself is like people don't necessarily want to upload stuff to cloud platforms because they don't trust AI companies because they're to train on the thing. Like we don't train on anyone's content. Like it's in our T's and C's, we're blue in the face from saying it. And like how we're addressing that now is we're actually going to build local versions of our technology. So you don't have to send audio to the cloud. We're putting our models that can run on your computer, do the mixing there. Actually you don't even need to, and the beauty of it is don't even have to spend time uploading stuff. I think. That's where we're going with on device, on-prem basically. So then people can work with unreleased material and there's no kind of fear of that. It's going to be trained. So yeah, that's where I think things are generally going anyway. Think music AI that we work with, they're moving that way as well. Like they want their models to be able to run on device essentially. So there's no, yeah, training, essential training going on.

Jakob Wredstrøm

Amazing. Well, David, it's so interesting to follow Roix again, as I said, I've heard about you guys so many times and LinkedIn, and obviously this is a space that I'm very interested in. And the more that I learned through technology, now Amplitude, we have around 40 developers and we have a lot of professors and academics and ML and AI. The more fascinated I am about... how you can actually really create these hyper specialized solutions and actually solve problems. And what we miss in the music industry a lot of times are these experts that are willing to take on these comparably small problems and market opportunities and just doing incredibly well. And we need these companies. And I think that makes music technology more interesting. So thank you for doing your part.

David

My pleasure. Yeah, really enjoyed the chat.

Jakob Wredstrøm

Great, David. We'll talk soon, right?

David

Yes, brilliant. Thanks so much.