AI Music Revolution

The View From the Curator's Chair

• Josh • Episode 29

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 17:56

Send us Fan Mail

Josh curates a SubmitHub playlist and gets thirty to forty submissions a week. From the gate, a pattern emerges that says something true about AI music: mastered AI tracks get approved at a high rate, unmastered AI tracks get declined almost every time, and the two are often the same song. The genre is won in the generation. The placement is won in the finish. This episode is the show's whole thesis, proven from the other side of the glass.

The Unlock System is JG BeatsLab's methodology for serious musicians working with AI tools. Lane 2 work: human-authored, AI-assisted music creation.

Visit JG BeatsLab: https://www.jgbeatslab.com

Every Thursday morning, I send out a newsletter. One email a week. It's where I share what I'm working on in the studio, the research running in the lab, and the methodology I'm refining in real time. New books, new Blueprints, new findings from the Red Lab Protocol research. If that sounds useful, you can sign up at jgbeatslab.com/newsletter. Thanks for listening, and I'll see you next Friday.

AI Music Revolution is produced by JG BeatsLab LLC, an AI music education company building the methodology, research, and community for serious creators working in Lane 2.

Get more from JG BeatsLab LLC:

Connect:

Contact: josh@jgbeatslab.com

Stop gambling. Start directing.

SPEAKER_00

Hello and welcome to the AI Music Revolution! I'm your host, Josh Galaleon, the founder of JGB Slab. Okay, enough theatrics. We got business to take care of. Today's episode comes from see, most people making AI music never get to sit into. But I get to, and that is one of a playlist curator. As you may or may not know, I run playlists on Submit Hub. And every week I get lots of submissions across the different playlists that I manage via Submit Hub. I'm going to talk about one particular playlist because I think there's a really good lesson here for everybody. So I run a country playlist. It's hard-rocking modern country with a real party vibe. And every week for that playlist, I get 30 to 40 songs submitted for it. And just to that, just that one playlist alone. And these are from people that I've never met. And what they're hoping is they're hoping simply that I put their track on my playlist because I have a lot of subscribers to my playlist. So on this one, I'm not the artist, I'm actually the gate for the artist. So I'm the person your song is trying to get past. And after doing this for a while now, couple years, a pattern has showed up in what I approve and what I decline that I can't unsee and I want to share. It tells you something about it tells you the truth about AI music. And I don't hear people really saying this out loud. So I'm going to say it out loud. We're going to look at what actually crosses my desk. The three stacks every week of submissions kind of that they kind of sort into, or three piles. And one comparison inside of those piles that should stop you cold when you're creating AI music. And then we're going to talk about two ways curators get this wrong. And the third position that almost nobody here is really talking about, even though it's the only one that gets you through the gate. And because this show is about getting real work done, we're going to land on exactly what that third position requires. This is the AI music revolution, so let's get into it. Alright, when I go through a week of submissions, they sort themselves into three stacks. Not on purpose, it just kind of happens this way every week, the same way. The top stack, the ones I approve the most, are AI-generated songs that somebody took the time to fully master. And I want to be really clear about this. My goal as a curator is to build the most enjoyable playlist. And so I'm not looking for AI songs to approve them. These are just the ones that tend to be the best. But I want to be honest with you about what those sound like because it honestly surprised me when I first noticed it. These songs sound awesome. They nail the genre, they hit the exact vibe the playlist is for. These are songs that are dialed into the audience. The sound is finished. And when it when it comes on, I think, damn, that's it. That's exactly what it is that I'm looking for. And those get approved at a very, very high rate. So again, these are AI-generated songs that somebody took the time to fully master. The middle stack is the human artist. These are real people, real instruments, real recordings. And here's the uncomfortable part that I have to be straight about. My approval rate for the human submissions is lower than my approval rate for mastered AI tracks. And this is something that is not on purpose. It often comes down to a couple things. One is the polish of these tracks. So you can tell that a lot of these human-created tracks were created by somebody in their home studio doing the best that they can with capturing their guitars and drum sounds and their vocals. But it's hobbyist. And it sounds like hobbyist. The other thing is a lot of times the song structures, they wander. It doesn't really have an identity of something that it's trying to be intentionally. It's kind of we're we just want to be a country song. And so they don't really hit that very specific lane that I'm curating for this playlist. And so that wandering sound, you just listen to it and it's kind of boring. Kind of like, where are we going with this song? Like, what's the vision for it? What's the mood that you're trying to sell me in the song? What are you trying to make me feel? What problems are you trying to take away in my moment right now? And a lot of times those those get missed in the human created tracks. And it doesn't mean that they're bad musicians because they're not. I hear a lot of excellent musicians that come across. It's just the songs are meh. And then there's the bottom stack. These are AI songs that were generated but not mastered. You can tell, you can hear it immediately. This is the raw generation. Somebody pulled out a sudo, they downloaded the WAVE file, and then they sent it straight to me, straight to Spotify, straight to the distribution, straight to me to review. I decline every single one of these. I'm sure that there's some that have probably gotten through, but the vast majority, if not all of them, I've declined. But here is the thing I really truly want you to sit with on this episode. Because when I noticed it, it really did reframe the entire thing for me. So the songs at the top of my stack that's and the song at the bottom of my stack are frequently the same quality of a song. They're the same genre accuracy. They have the same vibe capture. They've nailed the theme that I'm looking for. The unmastered ones that cross my desk often get the genre exactly right. The energy is exactly right. Everything in the generation is supposed to do, it did. The song underneath it is great. It's just, it's right there. And I decline it anyway. And the only thing that changed between the track I approve and the track I reject is the master. Same core song. One got finished and one didn't. And that one difference moved it from a near-automatic yes to a near-automatic no. So, really, when you when you think about it, I basically ran an experiment without meaning to. Because I'm running a business, I'm running a playlist business. But with this, with those two buckets of AI music, the song is constant. The master is the variable. And the variable decides really everything on the approval. Now, don't get me wrong, I have AI music that comes through that misses the point that I'm striving for, or just really isn't that good. So it's not to say that it's always an automatic approval because it's not. But when the underlying song is the same, but the master isn't there, then that's it. It's a no, it's not getting approved. That's the variable that decides everything. So this to me is really important for those of you who are trying to make top-tier AI music. Spend the time on the mastering. And this isn't this isn't just me having an opinion about mastering. This is what my approval logs show. The unmastered tracks that nail the genre still lose because they sound like exactly what they are. An export from Suno that somebody sent me before they actually finish the job. So there are two ways creators get this wrong, and both of them end up the same place on the wrong side of the gate. The first way is the one the bottom of the stack is full of. You fall in love with the generation, you get a take that nails the genre, it nails the vibe, and you think the song is done because it sounds done on a quick listen. So you export it and just send it out. And then I get it. And the output can genuinely sound impressive right out of Suno, but impressive on first listen and finish are two completely different things. And a curator hears the difference in about four seconds. The generation was the easy part. You stopped at the easy part. The second way is quieter. And it mostly affects the human artist and honestly, some of the people who have given up on AI entirely. It's the belief that the polish is out of reach. That sounding finished requires a studio, a budget, a real engineer, something you don't have. So you send me the track as is, the mix is decent, never fully finished, and you hope the song carries it. Sometimes it does, most of the time it doesn't. Because I'm sitting here next to a stack of tracks that did get it finished. And yours is getting compared to those. So both of those lose for the same structural reason. The gate is not judging your talent, and it's not judging whether or not you used AI. It's judging whether the track sounds finished. And hiding behind it's a good song loses to the finished track in the same inbox. And assuming the finished track is out of reach loses to the people who learned it wasn't. So the third position, the one people don't really talk a lot about, is the one that I think gets through. And the mastered AI track is my top stack. They're not winning because they're AI, they're winning because somebody finished them. Think about what it act what actually happened here. The generation nailed the genre, sure. The AI is genuinely good at that now. And it can hit a lane like the one I curate and get the sound right. We all know that. Give the tools due, that's real. And I'm not going to pretend it isn't. But nailing the genre is not what got the song placed. Half of my bottom stack nailed the genre too. What got the top stack songs placed is that after the generation, a human being took the track into a DAW and finished it. They balanced it, they glued it, they mastered it to the standard my ear expects when I'm deciding what goes on a real playlist. The genre was one in the generation. The placement was one in the finish. And here's why that should actually encourage you, whichever side of the AI argument you're on. The finish is a skill. It's a learnable skill. It's not talent you were born with. It's not a budget you need. It's not a record deal. It's a process you can run the same way every time. And the people beating you in my inbox mostly didn't out talent you. They just finished the job you stopped one step short of. So this this is the whole show in one sentence. And long time as listeners know it. AI moved the starting line, it did not finish the record. The curator's chair is just where you get to watch that be true from the other side of the glass. The tracks that finished the record get on. The ones that stopped at the starting line don't, no matter how good the start was. So let me be concrete about what the finish even means, because I don't want it to sound like a mystery only engineers understand. Once the song is right, finish it comes down to a handful of jobs. Balance, getting the levels sitting right so the vocal is clear and nothing is fighting it. Glue, making the parts feel like one record instead of stem stacked together. Loudness, bringing it up to the level streaming expects without crushing it flat. And the final checks, making sure it holds up on your phone, in the car, on earbuds, wherever, wherever you're going to actually listen to this. None of that is magic. It's all learnable. And all of it is the difference in my inbox between the top stack and the bottom stack. Now, if you want to learn that finish, the whole thing tone, glue, loudness, hitting streaming targets, turning a raw AI export into a master that actually sounds like a top top stack track. That's Unlock Reaper, Mastering AI Music. It walks you through building a professional mastering chain using free tools step by step, no expensive plugins required. It's $9.99 at jgbeatslab.com or on Amazon. And if you want everything, every book, the full generation to release system, the research, the blueprints, the sprint course, and fader, my AI studio manager that walks you through this work in real time, that is the Red Lab Library. Every piece of it is built the same way. One system, one price, $97 at jgbeatslab.com. So here's the through line from the curator's chair. It was never really AI versus human. It's finished versus unfinished. The tool got good enough to nail the genre. The finish is still on you. And the finish is the whole ballgame. So next time you get a generation you're proud of, before you send it anywhere, listen to it the way a curator will. Ask yourself one question: Does this sound finished? Or does this sound just impressive? Because the person on the other side of the glass is asking the exact same thing. So at the end of the day, it's pretty simple. Stop gambling, start directing. Thank you for listening, and I'll talk to you in the next one.