Transcript (auto-generated, may contain errors). Hosts are AI-generated. Stories curated by Synaptic Plunge.

Quick note before we start, we're AI-generated hosts and today's stories were curated by the Synaptic Plunge team. Welcome to Sculpting Sound. Glad to be here. Where can your AI songs actually live? Right, I mean that really is the question right now, isn't it? It is. Okay, let's unpack this. Imagine you're holding a freshly generated AI track, right? You've spent hours on it and you're just sitting there wondering which doors are about to open for you and And which ones are going to just slam shut?

Because you could have the most sonically perfect generation in the world, but if the platforms reject it, it basically doesn't exist. Exactly. So today we're taking a stack of sources to explore this rapidly shifting landscape of where AI music is actually allowed to exist. Yeah, there is a lot to cover.

We're going to look at the big streaming platforms, their walled gardens. Plus the fractured rules, the dependent sites. really crucial technique to make your track sound release ready. And finally, those fierce Ligel battles that could, uh, well, rewrite the rules for everyone. Yeah, that legal stuff is hanging over all of those. But to figure out where an AI track can live, we really have to start with the biggest player in the room. What is the top streaming platforms blueprint for the future of AI

music? Right, because whatever Spotify does, that sets the weather for the whole industry. It really does. And Spotify has a plan for an early 2027 launch of an AI-powered service, but it's not what people thought. They are allowing listeners to make their own remixes and covers of existing songs. Oh wow. So taking like a massive country hit and turning it into a dance track. Yeah, exactly. Or like a folk version of a big rock anthem. It's targeted as a premium add-on, so costing an extra four or five dollars a month.

Do we know if that's on mobile or desktop? The platform rollout is still unconfirmed, so so we don't want to guess there. But the business model is what's fascinating. Well, deals are already signed with Universal Music Group, Merlin and Believe, and negotiations are on going with Sony and Warner.

So the industry is buying in because it's controlled. It's like Spotify wants to turn every listener into a licensed DJ, but they're keeping everyone inside. a very of strictly monitored sandbox. The pushback online is interesting though. On X you've got people like user text for what Noah saying, you know, it needs to

to be fair because the song is someone's work, not just an afterthought for a remix. Which is a totally valid point. And then big Randy SR was questioning if the splits and the metadata for anyone who doesn't know is the digital paperwork,

the credits, the splits, the ownership info that travels with your file. If big music wants AI that is licensed and paid, your metadata has to be flawless. Because if fan remixes start paying out, missing a credit means losing money money.

The synaptic plunge team who actually makes original music for YouTube uses this as a daily reminder to keep clean records of every single release Titles, dates, prompts, lyrics, who made what, all of it Yeah, exactly. But if the biggest platform is focusing entirely on remixing the past, what does this mean for the listener who is using AI to create something entirely new?

Well, that takes us from the walled garden straight into the Wild West. Right. If Spotify is locking down their own catalog, where do your completely original, AI tracks go? It is a complete patchwork of rules depending on where you look. For example, SoundCloud is totally leaning in. 11 labs whose models have generated like over 25 million tracks. They now let users publish directly to SoundCloud via 11 Music. So you complete the download and reupload step? the Plyth step.

Exactly. It just goes straight there and gets labeled as AI. But then you look at Deezer and they are doing the exact opposite. Yeah, Deezer is actually adding a toggle to let listeners filter out AI generated music entirely.

Just wipe it from their feed. And Bandcamp just has an outright blanket band on AI generated tracks. Which is wild. And then you have platforms walking their own doors. The review site Litmus noted that since

an October 2025 settlement with UMG, UDO is basically blocked all audio, video, and STEM downloads. Wait, really? So you can't even get your file? Right. You can only share UDO links.

Their term state, the output is just for personal, non-commercial use. So you cannot post it to YouTube or Spotify or anywhere else. It's like trying to find a parking spot in a new city. But every single lot has completely contradictory signs.

It is so confusing for creators. And on X, people are frustrated. You have audio SAMC saying three platforms, three answers. And Derek Hoyam, questioning if a label actually changes how people listen. Although Osmoshua was really praising that 11 labs

direct to SoundCloud feature. Well, sure, because it removes friction. And standardizing this is the next big hurdle. DDEX is working on a Generative AI Declaration standard for early 2027. Okay, so a common way to disclose AI use on a release. But if you label a track, honestly, is AI, aren't you risking a listener hitting the

filter button on Deezer before they even give your song a chance? That is the fear. Yeah. But honesty actually protects you. If a platform tightens its rules later and you hid your AI use, you faced retroactive takedowns.

You could lose your whole account. Oh, that makes sense. Yeah, disclosure builds audience trust, and it protects your discovery, meaning how new listeners actually find your music in the long run. You definitely have to pick your platform and check its specific AI policy before planning your release.

Always label honestly. The Synaptic Plunge team, they publish their albums on YouTube and disclose the use of AI up front. Which aligns perfectly with where the platforms are heading anyway. Disclosure or over hiding? Okay, here's where it gets really interesting.

Once you've navigated the platform rules and found a home for your track, you have to make sure actually sound good enough to compete. Right, because if you make it past the rules but the song sounds like a robot in a tin can, no one's listening. So how do you polish an AI track to standard without ruining the original file? Well, the Suno studio had these big web updates on October 29th.

They brought in fine pitch adjustments and auto stretch for added audio. Auto stretch is a game changer. It really is. The final pitch adjustment is what we need to focus on. We have to define some terms here. A semitone is the jump from one piano key to the next.

Which is a pretty big jump. It is. A scent is one hundredth of a semitone. And when we talk about fixing audio, we have to understand time stretching, which changes speed without altering pitch, versus pitch shifting. Right.

If you pitch a human voice too far, you mess with the formants. Okay. Wood or formants? Formants are the natural voice resonances. the physical size of the throat. If you shift them too far, you get that classic chipmunk effect. I hate when that happens. Instead of tearing down the entire house because one door squeaks,

you just oil the hinge. That is exactly the technique. Make small moves. We're talking one to three semitones max, or just 10 to 50 cents for fixing sour notes. And you have to work on one stem at a time. A stem being the separated individual track like just the vocals or just the drums. Right. Because if you shift a full mix, it confuses the algorithm and smears the drums. You start hearing these awful metallic or underwater artifacts. It sounds so cheap when that happens.

So you do your pitch changes on the stem, and then you do your EQ and compression after the pitch changes. Does this mean the AI music workflow is finally starting to look exactly like traditional audio engineering? It really does. It's all about the physics of sound now. Working in tiny fractions of ascent works so much better than the sweeping drastic changes.

And the pros are noticing this. On X, Blackout the CEO was loving the fine pitch for tightening vocals. Track Finish Lab was praisinging the workflow shift. Reworking parts instead of re-rolling the whole track. But they did warn about bleed between stems. Bleed is when a little bit of the drums gets stuck in the vocal track. Right, which ruins the pitch shift. But VBN Studio.ai had a great tip for that? Ask for sparse arrangements in your prompt so the stems separate cleanly later. That is incredibly smart.

Fixed tuning with tiny scent nudges on one stem at a time, rather than throwing out a whole song. This synaptic plunge team actually uses this precise method to save credits and keep the takes they already like. It is the most efficient way to work right now. But, of course, hovering over all of this, making the song, tuning it, choosing a platform is a massive legal cloud.

Yeah, we have to talk about the lawsuits because they That is a mountain of paperwork. It is. And NVIDIA says they shouldn't have to provide private opinions just to act as expert witnesses for the labels. And Suno is pushing back too, seeking extreme volume data from the mechanical licensing collective. But the MLC is fighting that because of the massive staff time it takes. The legal stakes here are monumental. The labels are trying to prove firm. They want to show AI hurts their business.

Because that would defeat Suno's fear of use defense. Exactly. No, it's intense. On X, vocal AI was breaking the news on NVIDIA, and Moses Muse pointed out something really important. Soono-Ink is the defendant, not individual users. That is a crucial distinction. Real Khan was also pointing out that the labels are trying to frame all the AI output

as product from the poisoned well. So what does this all mean for you, the listener? If I'm just making a crack in my bedroom, why should I care about what NVIDIA's venture arm is fighting over in court? the outcome of that fight, fair use versus market harm, will completely dictate the future licensing of every single tool you use. If the labels win, the cost to use these generators could skyrocket, or they might all end up

locked down like Udio. But for now, nothing changes in your app today. The lawsuits target the company, not you. However, you absolutely must keep meticulous records. Dates, prompts, original lyrics, subscription plans. Just keep them all safe while this plays out.

The Synaptic Plunge team actually keeps original lyrics and prompts with every project just as cheap insurance. It is the smartest thing you can do right now. Document everything. Okay, so to recap the four keys to where AI music can live. 1.

Prepare for Spotify's licensed paid remix future by keeping your metadata totally clean. 2. Check the independent platform rules before releasing. the difference between SoundCloud and Bandcamp and label everything honestly. Three, polish those tracks using microtuning in sense and do it on individual stems. And four, keep all your creative receipts while the courts figure out these massive legal precedents.

It's a lot to navigate. It is. And it leaves us with one final thought to really mull over. If platforms eventually require a verified generative AI declaration for every single track, will Will the definition of original music permanently split into two entirely different industries? That's a wild thought. Or will the line blur so much that, you know, ten years from now no one cares how a song

was made? Only how it makes them feel. 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 Grocking It.

See you in the next one.