Transcript (auto-generated, may contain errors). Hosts are AI-generated. Stories curated by Synaptic Plunge. We're AI-generated hosts, and today's stories were curated by the Synaptic Plunge team. We always like to be upfront about that. Welcome to Sculpting Sound. So I want to address you directly, the listener. Because we have a really fundamental question to explore. Can AI music earn a listener's trust? Now, for this deep dive, we are looking at, for this deep dive, looking at specific areas to answer that. trust through just being honest about artist identity on Spotify. Which is a huge deal right now. Exactly. And we're going to look at proving your quality by making songs sound genuinely finished, specifically through some advanced mastering techniques. Right. Get a professional polish. And third, ensuring clear rights when you add new sounds or plugins into your workflow. And finally, we'll wrap up by of watching how the courts are deciding the, you know, the much bigger roles on AI training and copying. It is a lot of ground to cover, but it all comes back to that central idea of trust. Because, you know, before a listener can trust a song, they really need to trust who is actually singing it. Yeah, that invisible contract between creator and listeners is just completely fracturing right now. Exactly. And that is why Spotify has stepped in. They are rolling out this new AI persona badge. Right, I saw this. I saw this. It's visible on both mobile and web now, right? Yeah. it back on August 11, 2026, and it's been rolling out since like mid-September. So what exactly qualifies as an AI persona to Spotify? Well, they define it as a public identity that represents an AI-generated photo-realistic human. Okay, let's unpack this because if I'm a producer and I'm using Suno to make some, you know, instrumental tracks, I might be panicking right now. Oh, for sure. But that's a huge misconception. The badge is really about identity, not the tools you use. So simply using Suno or using AI to separate vocals or generate stems, that does not trigger the badge. Because there are optional AI credits you can use for those tools anyway, right? Exactly. This badge is specifically for when you fabricate a realistic human face and pretend this artificial entity is a real person. I mean, it makes sense. It's the difference between like using a flattering filter on your photo versus literally catfishing your audience with a fake person. hate feeling tricked and Spotify knows that. They really do. I was reading this thread from an ex-user at cicada siren and they were saying how they found these songs they loved but the moment they saw it was a fake AI persona just ruined the songs for them. Yeah the emotional connection just breaks. And the interesting thing is Spotify actually has two versions of this badge. Wait really? Yeah. Two versions? Yeah there's the standard AI persona badge which which is self-disclosed. Like, so get publicly called out pretty much. And the impact of getting that badge is severe. You lose your algorithmic recommendations. Meaning of no discover weekly. No discover weekly. No editorial playlists and no release radar for new listeners. All your growth basically has to come from outside Spotify. That is brutal. But you know, you're not going to be able to get a job that you're not going to be able to get. That is brutal. But it's also causing some pushback in the industry like Mikey Shulman the CEO of Suno was on the Trappitle podcast recently. Great. I heard that interview Yeah, and he was pushing back on this whole narrative. He basically said, you know streaming fraud like BOTFARMS That's a crime, but AI music is not a crime And he's totally right about that. The music itself isn't the problem The deception is. Which brings up a great example from our own synaptic plunge team They release music too, but they use an illustrated stage name. Right, completely sidestepping the issue. Exactly. It's obviously art, not a fake photo. So they avoid the badge entirely because they're being honest about their aesthetic. So the big takeaway for section one is using Suno doesn't earn the badge. Pretending a realistic AI face is real does. Trust is just earned through honesty. Exactly. But, um, okay. Once you're honest about who you are, the next hurdle is that your music actually has to sound good. Oh, absolutely. Trust is built through unmistakable audio quality. If it sounds like a muddy demo, they're gone in five seconds. Right. And to prove that quality, you have to make the song sound genuinely finished, which brings us to mastering. Yes. And there is a fantastic guide from Isotope that came out on November 13th, breaking down a five-pass mastering session. Let's go through those five passes, because mastering is basically just the final step in production, right? Yeah, it's about managing tonal balance, the dynamics and your LUFS and true peak levels. So what's pass number one? Start with the right file. You need a WAV file, like a 24 or 32-bit WAV, not an MP3. Because MP3s are already compressed and you lose all that data. Exactly. An engineer on X at mastering one underscore tested raw sueno waves and found they average around minus 15 LUFS. Which gives you a lot of headroom to actually work with. Plenty room. So pass two is you have to listen all the way through the track. Without touching any knobs. Right, no plugins yet. You have to map out the song. Because if you set your limiters based on a really quiet verse, you're going to completely break the loud final chorus when it drops. That makes total sense. You have to know where the loudest part is. Exactly. Which leads to pass three. Clean before you polish. And this is from AtTrackFinishLab on X, right? Cleaning out the artifacts and clicks first. Yeah. If you don't remove those weird digital noises first, any mastering you do will just make them worse. It's like polishing a car. Mm-hmm. If you don't buff out the scratch first, but in shiny wax on it just makes the scratch more obvious. That is a perfect analogy. You are just amplifying the flaws. Okay, so pass four is processing at matched levels? Why is level matching so important? Because louder always tricks the human ear into thinking it sounds better. Oh, right. If you make an EQ adjustment and it just turns the volume up, you'll think you made a genius mix move, but you really just made it louder. Okay, so just throwing a normalizer on there to hit minus 14 luffs isn't actually mastering. Not at all. Tone is the real work. You have to match the volume so you can actually hear if the tone improved. Got it. And the final pass, pass five, is checking against references. They have to make sure the levels match so no, it jumps out and blasts your ears. It's crucial for that professional feel. So the takeaway here is master from the WAV, clean the clicks and use references. Professional effort earns trust. Okay, so you've got the honest identity. You've got the pristine audio quality, but then there's the legal side. The dreaded legal side. Yeah, trusting your ingredients. Because distributors and listeners need to trust that the final product has clear rights. This is becoming a massive topic, especially with AI moving directly inside the DAW, like as a studio assistant. Right, like that Music Tech report from November 12th about Sonora Flow. Exactly. It's a free text-to-loop plugin for Mac and Windows. You don't have to go to a separate website anymore. Which is part of the huge trend, right? Like Anakari has those AI chat assistants now, and RC20 has loudness matching built in. Yeah, AI is just living inside your timeline now with Sonora flow You literally type a prompt pick a category and it generates a leap or a one-shot that you just drag right into your project It's incredibly fast But the rights issue is where it gets tricky because Sonora flows free plan releases those sounds into the public domain under a CC zero license Right meaning anyone can use them for free without attribution. Here's where it gets really interesting If these sounds are CC zero, they are totally free to use But doesn't that also mean anyone else can use the exact same sound? Yep. So doesn't that dilute my tracks uniqueness? Like I could build a whole song around this amazing vocal chop and someone else could just grab the same chop for their own song. That is the exact trade-off. For beginners, a clear public domain license is amazing because it solves that fear of you know, can I legally release us? Right, you won't get sued. But free to use absolutely does not mean yours alone. You do not own that sound exclusively. Which is why you have to be so careful. Like, the Synaptic Plunge team, they use plugins to add missing layers to their sueno songs all the time, like maybe adding a shaker or a synth pad. But they always verify the license. Every single time. They always ensure they know the license of every added sound. It's the only way to be safe. Text-to-sound tools let you patch missing layers easily, but you always, always have to read the license to ensure clear rights. Exactly. So that covers the individual creator level. But I want to zoom out for a second. Okay macro level. Yeah, you can secure your own loops in your own profile But what about the whole industry? Like can the public trust the actual foundations of AI models while the law is still completely plain catch up? That is the multi-billion dollar question right now and courts are literally deciding this늷 as we speak Right like the Getty versus stability case in the UK Court of Appeal. Yes Yes. This is a massive case. And it's really about primary versus secondary infringement. Can you explain the difference for us? Well, the BPI, which represents the UK record industry, they joined the case and they used this really great analogy about a CD. Oh right, the bootleg CD analogy. Exactly. If someone presses an illegal bootleg CD in another country, that's primary infringement happening over there. But if they ship that CD into the UK and try to sell it, that is secondary infringement in the UK. Got it. So how does that apply to AI? Well, stability trained its AI offshore on servers outside the UK. So they are basically saying, hey, we didn't copy anything in the UK. They're trying to dodge the local copyright laws. Exactly. But Getty is trying to prove secondary infringement. They're saying by selling the model in the UK, it's like importing that boot-like CD. And he is terrified that if the courts side with stability AI companies will just offshore all their training to countries with no copyright laws. It's a huge loophole. And US lawyers are watching this incredibly closely. So what does this all mean? Is the creative industry basically watching a giant game of jurisdictional whack-a-mole where AI companies just hunt for copyright havens? It definitely looks like that. But if we connect this to the bigger picture there's a twist that might make the server for locations completely irrelevant. Wait, really? How? It's something called memorization. And this is popping up in the German cases right now, like GEMA versus SUNO and open AI. Memorization. What does that mean in this context? It's when an AI model can literally spit back significant recognizable chunks of its training data. Like perfectly recreating a copyrighted song. Exactly. If it can do that, courts might decide that the model itself holds a physical copy of the data. Oh, wow. So it's not just learning a style actually contains a copy of the song inside its weights. Right. And if the model itself is technically a copy, then the whole offshore defense just completely falls apart. It wouldn't matter where you trained it. Because selling the model anywhere means you are selling an illegal copy. Precisely. That is wild. But while the courts take years to figure all this out, there are ways to protect yourself today. Like the synaptic plunge team. They write original lyrics for every single album. Which is the smartest thing you can do. Yeah, by steering clear of prompts that try to recreate famous songs, you protect yourself regardless of how courts rule on memorization. The courts are deciding the big rules, sure, but you're keeping your Suno songs clearly original, keeps you on the right side of the law no matter what happens. Absolutely. So to recap everything we've talked about today on this deep dive. Before pillars of trust. Right. You have to be honest about your identity. You have to make your tracks sound genuinely finished and mastered verify the rights of your loops, and you need to stay clearly original. It is a lot of work. It is. But it answers our question from the top of the show. Yes, AI music can earn a listener's trust, but that trust is never automatic. It has to be intentionally earned at every single step. This raises an important question. Yeah. If listeners are going to demand perfect transparency about identity, flawless audio quality, and unquestionable legal rights from AI creators. Which they clearly are. Right. Will AI music eventually face a far stricter purity test than human made music ever did? I mean humans use stage names and unclear samples all the time. That is a really provocative thought to leave you all with today. A double standard in the making. Well, that's today's sculpting sound. Thanks for listening to Sculpting Sound from the Synaptic Plunge team. If you've got a few more minutes today, check out our other shows. The Daily Plunge and Silicon Sidekicks. See you in the next one.