Transcript (auto-generated, may contain errors). Hosts are AI-generated. Stories curated by Synaptic Plunge. But today theo h Right. So the mission for today's Deep Dive is to basically teach you how to build a motor for that little papal boat. We've got five crazy stories to get through. And we really need to start on the front lines, which is the micro content layer. Yes. Because if we analyze how music actually spreads today, it doesn't start with a full album drop on Spotify. You know? No. Nobody does that anymore. It starts with a 30 second vertical video on social media. That is your foundational node for everything else. It's the anchor. And if you don't control that initial 30 second hook, I mean your downstream revenue, your algorithmic traction for the whole song, it's all fundamentally compromised. Which brings us to our first story. Meta is rolling out this new content protection tool for Instagram. And they are essentially handing you a control panel. Yeah, and to really grasp the significance of this, we have to look at the architecture Meta is using. For years, the major record labels have had the system called Rights Manager. The big enterprise level stuff. Exactly. It's this highly sophisticated digital fingerprinting environment when Universal or Sony uploads a track, Rights Manager creates this complex cryptographic hash of both the audio and the visual components. So it's not just looking for like the file name? No, not at all. It analyzes the actual frequency spectrum and the pixel distribution. And what's wild right now is that Meta is taking a streamlined version of that exact enterprise architecture and just embedding it directly into the professional dashboard of the everyday creator on Instagram. Which completely shifts the power dynamic. I mean, think about the compute power needed for that. It's staggering. They are scanning billions of uploads across Facebook and Instagram basically in real time. So picture this. You make a new song in Suno. You take a 30 second clip of the chorus, put it over your album art and you post it as a reel. your original anchor. Right. And then a week later, some aggregator account rips that exact video, posts it, gets 40,000 views and gives you zero credit. Which happens constantly. Constantly. But now, the system doesn't just guess. It runs a deterministic, your original hash. And this is where the interface design gets really fascinating psychologically. Because Meta isn't just giving you a binary takedown button. No, they give you choices. Right. Four of them actually. When it flags a match, you get this dashboard showing the analytics of the stolen copy. You see the reviews, their follower account, and you can choose to just track it, which is the default. Kind of just gathering intel. Yeah, seeing how your music performs in their algorithmic cluster. Or you can hide it, which cuts off its reach without penalizing the other account, so you avoid the whole platform relations drama. Or you can release the claim, or, and this is the big one, you can add a credit link. I really want to dig into that credit link because from a strategy standpoint, it is a brilliant judo move. It really is Like your first emotional reaction to seeing someone steal your work and get 40,000 views is rage You just want to nuke the post. Oh, absolutely. You want it gone? But economically Zering out 40,000 views is basically destroying potential capital If you use the credit link instead You retroactively co-opt their distribution network You turn the thief into your unpaid marketing intern. Exactly. The platform slaps this permanent original buy tag on their video that funnels their viewers straight back to you. You get free promo. But wait, I have a question about this. What if they don't steal my video but they just steal my voice? Ah, yeah. That is the glaring vulnerability and standard visual matching. Because if someone rips the audio of me talking, strips my video, and plasters a fake AI I generated face on top of it. The hash won't align. The pixels don't match anymore. Right. So what happens then? Well, this is where Meta is attempting a massive technological leap. They're actually starting a small test to catch reused spoken words from original videos, completely independent of the visual anchor. Wait, really? That's insane. It is. Think about the machine learning required for that. The system has to decouple your phonetic cadence, The unique spectral signature of your voice matched that against the database, all while actively ignoring the fake synthetic face on top of it. That's a direct computational assault on AI cloning. It really is. But it steps into this incredibly complex legal gray area. We really have to separate platform policy from actual statutory law here. Right, because clicking hide on an Instagram dashboard isn't a federal DMCA takedown. Exactly. This is just platform policy. The reason Metta is accelerating this spoken word test is because global laws are just failing to keep up with generative AI. Copyright law protects fixed expressions like a recorded song. It struggles with abstract concepts like the likeness of a voice. Which usually falls under state by state publicity laws, right? Yeah, it's a mess. So the platforms are basically acting as pseudo law makers because the actual law makers are paralyzed. And on X, people are talking about this. sure how Fixer was pointing out the eligibility requirements for the dashboard. And the white page EN was tracking how the rollout is actually flagging these copied videos. So the big takeaway for you here, if you post song clips on Instagram, check for that content protection tool in your professional dashboard. It's rolling out on Android first, probably iPhone soon. Make sure you post your originals first and definitely use that credit link. Yeah, build that anchor. Because that distinction between platform policy and law, perfectly transitions us into what happens when these systems fail catastroph history and tech years. They assumed good faith. They assumed a human-to-human dispute, or at least a corporation to human dispute. The idea of an entity generating thousands of synthetic tracks a day and firing off automated takedowns was science fiction. But now it's bought against human, and the human is losing. Let's look at the concrete example here, the Bray-Patrunko incident. Such a vital case study. So picture this, you have this Bulgarian folk song from the 19th century. Definitively in the public domain. who owns the composition. A human musician records a 50 second, entirely acapella version of it. She provides all the physical energy. Exactly, her vocal cords, her microphone. She uploads it to YouTube. Four years go by, it gets 125,000 views, and then boom. An entity called Media Nov Records triggers a copyright strike, and YouTube riffs the video down. And when you look into Media Nova records, It gets so dystopian. Based on an investigation by 404 Media, this company boasts about AI-powered content recognition, but structurally, they're running about 500 YouTube channels. 500 channels. Just churning out content. Yeah, these high-volume content farms, just pumping out repetitive electronic music and AI imagery. And to protect their territory, they run automated scripts that scrape the platform and issue claims based on incredibly loose algorithmic matching. Here's the human singing the public domain melody and because its own AI slop was trained on that same melody It thinks the human is stealing from the bot exactly the system penalized the human origin to protect the machine output And what's crazy is that while the human was blocked her actual Organic voice was still living on in all these popular AI remixes across the platform That is infuriating and the asymmetry of effort here is the real issue right for the bot a claim costs a fraction of a cent in compute power. For the human, you have to navigate convoluted legal forms, wait for platform review periods, and your appeal basically just goes straight back to the bots on red inbox. And in this case, the creator sent emails, filed two challenges, all disappeared into the void. It only got resolved because a human journalist from 404 media reached out for comment. And minutes later, the company withdrew the claim saying there was no revenue or views, which was was a blatant lie, it had 125k views. So if you're an indie creator making songs with Suno, how do you protect yourself from a bot army like this? Because legally the DMCA has a section, Section 512F, that that says false takedowns carry liability for damages. Yeah, on paper it does. But practically, it's useless for the indie creator. Enforcing that means hiring a lawyer and going to federal court. You're talking tens of thousands of dollars to fight a faceless LLC over a video making five bucks a month. The economics of justice just completely insulate the bot farms. Exactly. As user SISpotlight pointed out on X, these churn channels operate with impunity. And Zhuo P. Matias asked the Million Dollar question. What proof should YouTube actually require by a video down? Until they fix that, your defense strategy has to be proactive. You can't rely on reactive legal action. You need the synaptic plunge defense routine. Yes. It is entirely about your digital paper trail. First, keep your receipts. Keep the Suno Songpage link, dates, lyrics, stems. Second, you have to upload your version first to your own channel, with a clear description stating it's AI-assisted. You're basically manipulating the ingestion timeline. Precisely. You have to beat the bots to the database. And check your YouTube Studio dashboard fast and frequently on your phone or web. If a claim hits, dispute it immediately with specifics. Say you own the sound recording arrangement. that paper trail. So upload first, keep proof of how you made it. That is your strongest shield. But all this administrative warfare overwrites kind of assumes someone is actually going to listen to the song once you secure it, right? Which brings us to the most depressing metric in the modern music industry. Yes, section three. The void of attention. Because you fought off the bots, your song is online, your control right, wrong. You cannot control human attention and and the numbers are just staggering. The scale of the digital archive has completely outpaced our biological capacity to listen to it. I wanna drop this bomb because I couldn't believe it when I read it. Across streaming services, Spotify, Apple, all of them. There are roughly 300 million tracks live right now. 300 million. Out of goes, only 40-million have ever had a single stream. Wow. that leaves 260 million uploads with zero plays. Zero? Not 10, not 100, zero plays. That's roughly 87% of all music online. And this data was reported by Digital Music News, corroborated by a top distributor exec. Deezer also noted that fully AI generated tracks make about one to 3% of streams, but 85% of those were flagged as fraud. So uploading music today isn't like stocking a shelf in a store. It's like throwing a single grain of sand onto a massive beach. Volume does not equal attention. It doesn't. And we have to look at the algorithms to understand why platforms rely heavily on collaborative filtering. The algorithm doesn't listen to a song to see if it's good. It looks at human behavior. Right. if I like a song and you like similar music, it recommends it to you. Exactly. But it needs an initial seed of human data to do that. If you just dump 100 Suno tracks onto Spotify with no context, the algorithm has no idea who it's for, so it just sinks to the bottom. Which explains why these virtual AI personas are completely losing steam. Oh, completely. The data shows virtual artists are collapsing. Agvers dropped from 257K listeners to under 50K, an 82% drop. Sirewell fell 88%. Spotify actually nuked 75 million spammy tracks and pulled 50 million AI tracks from their auditions. People don't want purely synthetic artists. They want a parasocial connection. And honestly, this is secretly good news for real creators. AI Slop isn't drowning out real music because fans are actually picky. Yeah, they want a narrative. But there is a crazy legal angle to these zeros. Suno is actually pushing to subpoena stream volume data from the mechanical licensing collective, the MLC and from Apple. Why would they do that? It's a fascinating legal maneuver. It is. The labels are suing Suno for copyright infringement, right? Saying their tech causes market harm. , but if Suno can prove in court that 87% of AI generated songs get literally zero streams and make zero money, they can argue there's no market harm. Exactly. They're basically saying, look, our tech isn't stealing Taylor Swift's market share because nobody's even listening to our outputs. That is a profound legal paradox. Defending your tech by proving it's commercially irrelevant. But for the indie creator, the lesson is intensely practical. You have to stop treating releases like lottery tickets. To beat that 87% void, the synaptic plunge methodology says, release fewer, better songs. Tin finish tracks are better than a hundred rough generations. them a home like a YouTube video, tell a story, give people a reason to listen. Be totally open about using AI to build trust. And whatever you do, do not try to game the numbers with bots because that just leads to takedowns. Treat every release like a real launch. But okay, so we talked about the cloud, we talked about distribution. Let's move into the control room itself. The DAW. Yes, section four. Because all those 260 million un-undered exist because AI I made generation easy. But now AI wants to be your assistant engineer. Your digital intern. Right. We've got chat boxes inside our DAWs now, like Apple's Logic Pro 13 on the Mac and Suno Studio's chat feature on the web. But should you actually let them make your track? Well, let's compare them. Logic Pro 13 has this intelligent assistant where you use plain language commands, you ask it to find regions or match tempos. Suno Studio has a chat for bulk actions, like renaming all tracks based on the clip name or bringing all faders down by 3DB. Which sounds great in theory. Then you read the CDM review of Logic 13 and the reviewer literally called the presentation apologetic. Yeah, that's not a word you want associated with a pro audio tool. No, so I have to ask, is it actually helpful or is it just the digital equivalent of an intern spilling coffee all over the mixing desk? It leans toward the coffee spill, honestly. The problem is a fundamental clash between natural language processing and deterministic software. A DAW is a precise mathematical grid. A chatbot is just a text predictor. So when you ask it to match tempo to a stamp, it tries to translate that into API calls and it fails. The reviewer said matching tempo caused chaotic, localized tempo changes that broke the whole project grid. Simple jobs are slower. Creative requests feel like a randomizer. And for basic help, it just dumps text from the manual at you. And the monetization is so frustrating. Logic is usually a one-time purchase, which is great. But this assistant is locked behind a new 12.99 a month Apple Creator Studio sub. While genuinely revolutionary features like the neural� network stem splitting are free in the same update. It feels like they just bolted a chat bot on to please shareholders. The community reactions on X validate that. CD-M blogs questioned if this tech is even the right solution, while JCM app studio and Huli sessions were super frustrated by the paid tier. Great Coaches was just like, is this even worth the money? And legally, there is a massive authorship question here. Oh, right, the Zari of the Dawn precedent. Exactly. The US copyright office ruled that prompting an AI does not equal human authorship. So if you use a DAW assistant to rename tracks, you're fine. But if you tell the chatbot, write a B section with a minor chord progression, and it does it without your manual input. You might be creating an uncopyrightable slot project. You're stripping away your own legal protection. So our advice, let the bot do the chores. Cable management, track naming, leveling. Keep the creative calls, the sound selection, the arrangement for yourself. Never delegate your creativity to a chat box. Which AI generated song might actually be a tool that completely ignores the AI trend. Ableton Live 13. Yes. Currently in public beta, releasing early 2027. And it is desktop only, Mac and Windows. And what's wild is their headline features, zero AI. Zero. None. Instead, they are focusing entirely on tactile hands on control. They're adding AR2 support, which comes bundled with Melodyne Essentials. They have a massive new 128 step step sequencer built right into the clips. That sounds incredible. A huge new drum synth spanning classic models to physical bongos and racks that split by frequency, like mid-side processing. It's such a contrast. It is. Apple built a chatbot, Ableton builds a sharper scalpel. And the CDM reviewer noted that just the preset navigation in Ableton made him, quote, switch and not look back. But we have to highlight AR-A2. This is a game changer. Explain that for people who aren't familiar. Historically, a pitch correction like Melodyne had to be inserted as a plugin. You had to play the track in real time, record it into the plugin's buffer, and then edit. With ARA2, the plugin shares the host memory. You see the spectral data instantly on the timeline. No real-time transfer. So if I'm making music with Suno, why do I care about a DAW with zero AI? Because the Suno song is just the starting point. Live 13 helps make it yours. Think about AI stems. They often have micro errors, right? A vocal pitch drifts slightly or the rhythm is stiff. Yeah, the generation artifacts. Right. With ARA2, you drop that sueno stem into live, use Melodyne right on the the timeline and surgically fix those drifting notes. You apply human, deterministic correction to machine output. And what about the drums? AI loops are always so lifeless. You replace them. Use the new step sequencer in DrumSynth. You can add a ghost snare with probability chance settings, it injects that human unpredictability that a rigid AI loop completely lacks. And then for shaping the sound, if the AI gives you this dense, muddy block of audio, you use the multi band racks, add drive only to the guitar's mid range, keep the sub clean, you manually untangle the sound wave. The community is going crazy for this. On X, Ableton confirmed the 2027 release, an infinite strategy praise the 14 headline features specifically saying not one of them is AI. Sidebrain said he's going absolutely mental over the features. Because producers want better editing tools, not software that writes the song for them. Exactly. So the best tools for making an AI song your own aren't AI at all. Let's recap this whole supply chain we've mapped out today. Let's do it. Step one, protect your micro content on Instagram using content protection and the credit link. Step two, build a proactive paper trail on YouTube to fight off automated claim bots. Step three, beat the 87% zero stream void by releasing fewer songs with actual narrative intent. Step four, delegate the boring chores to da chat bots but keep the creative decisions human. And step five, use highly tactile tools like Ableton Live 13 to surgically edit and inject human variants into those AI stems. So to answer your title question, who is really in control of your song now? Right. but only if you actively claim it. From the very first sueno-prompt to the final melodyne edit, control belongs to the person who leaves a definitive human fingerprint on the entire process. I love that. And I want to leave you all with a final thought to chew on today. Think about the music that you genuinely love. You don't love it because of how fast a server rendered it. You don't care about tariff-lops. You love it because of the deeply human choices the artist made, the slight vocal hesitation, the weird snare tone, the vulnerability. The bots have already won the war of volume. Your job now isn't to out-generate them, it's to out-human them. Absolutely. 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.