Transcript (auto-generated, may contain errors). Hosts are AI-generated. Stories curated by Synaptic Plunge. Imagine you discover this brand new song on your weekly discovery playlist. Oh, I love when that the singer, the instruments, the emotional inflection, it was all generated by a server rack in a data center in about three seconds. That is such a jarring realization. It really is. So do you still love the song? Does it change the way you value that experience? I mean, that's the ultimate question we're facing right now. It is. And according to a massive new global 80% of you want to know the truth before you even press play. Welcome to today's Deep Tive. I'm your host. And I'm here with our resident industry expert to explore a fascinating stack of internal documents news and survey data. I'm glad to be here. And yeah, we have a ton of material to get through today. We really do. Everything we are looking at today coincidentally centers around one specific flashpoint date, which is November 17, 2026. It was a wildly busy day for the industry to put it mildly. Oh, completely. And our mission today is to pull all of these threads together to help you understand this rapidly shifting landscape of AI music. it is moving incredibly fast. Exactly. And hanging over all of these leaps in tech and these big courtroom battles is one central defining question. As the lines between human and machine blur, should AI music come with a label? I mean, it is the defining tension of the modern creative era right now. Absolutely. We're looking at a landscape where the tech has evolved so aggressively that traditional boundaries of music creation and global copyright Law are bucko under pressure, they can't keep up. Right. piece of source material. It's an internal format document from an independent AI music the music show called Sculpting Sound. Oh, right, produced by the Synaptic Plunge Team? Yep, the Synaptic Plunge Team. This is a document strictly mandates that every single episode has to open by stating two specific things. First, that the hosts are AI-generated voices. And second, that the stories are human curated. Exactly. It's a completely non-negotiable opening for opening for them. Okay, let's unpack this. Think of this internal document kind of the ingredients label on a can That is a great way to look at it. Front. Yeah. Saying here's the artificial flavoring and here's the human o-sourced nutritional value. Value. And noting how this sets the stage for our entire deep dive is crucial because the whole industry is currently wrestling with whether that kind of transparency should be a choice, a feature or an actual legal requirement. Right. Because relying on voluntary labels is a pretty dangerous game on the internet. Oh, totally. So if transparency is the goal, how is that being built directly into the tools creating this content. Let's start at the Genesis point which is the creation platforms themselves. Which brings us to the big November 17th launch. Yes. Suna launched their huge new visuals feature that day. They rolled it out simultaneously on Android, iPhone and the web. It was a massive pivot for them honestly. Prior to this they were just an audio engine. Yeah, but visuals turns existing Suno audio library tracks into these 15 second video clips. which is perfectly optimized for things like shorts or reels. Exactly. And it does lyric videos, custom album art, and it even even features lip sync to the generated vocals. Plus you can re-prompt it with just plain text. Right. So if you generated a funky road-chip song, you can literally just type convertible driving through a desert at sunset cutting on the beat. And it actually forces the visual transitions on the given the server load. And competitors are racing them, right? Oh yeah. Indie AI Maker TuneArt is using cling and sea dance models to do the exact same thing. It is a race to see who can turn finished songs into finished videos the fastest. What's fascinating here is the strict guardrails Suno put in place. Yeah, they are not missing around with that. No, they actively block prompts that request specific real artists, public figures, or even copyrighted characters. So you can't just generate a video of Taylor Swift singing your AI song. Exactly. The system just bounces the prompt. And before a video can even be shared outside the Suno app, they apply an embedded AI watermark. Which is huge. That is labeling built directly into the source file. It's permanent. Right. And the social reactions on X were immediate. You had at Aang Girlfriend calling it a really fun product to work on. Yeah. And at Michael Ganadi was just expressing pure shock at the tech. And at Rafe AI Tech noted that it literally brings sound to life. game changer for indie creators. But I want to push back on that 15 second limit for a second. Why cap it at just 15 seconds when songs are normally like three minutes long? It does seem pretty restrictive at first glance. Right. I. I compared to the early days of silent film moving to talkies. Right now, AI video is just incredibly computationally heavy. Oh, the GPU processing required is astronomical. Exactly. So the 15 second hook is realistically the most economical way for them to dominate social media feeds without bankrupting their servers. That makes total sense. It's an economic necessity disguised as a feature. So, Suno embeds a watermark on the video. But what happens when bad actors strip that away, or just bypass video entirely, and impersonate an artist's audio to get on major streaming platforms? And that pivots us to the distribution battleground. Yes, Spotify. Also on November 17th, they rolled out this new artist profile protection feature to all artists via the Spotify for artists dashboard. Available on the app and the website for phones and desktop. Right. And it combats three specific threats. Honest metadata mistakes, bulk spam uploads, and then the big one, AI impersonation. Which is where someone generates a fake track and literally tags a real artist in the metadata. So how it works is artists get these alerts when music is delivered under their name. And they can remove themselves to prevent it from hitting their profile or the algorithm. It acts as an early warning system basically. Here's where it gets really interesting though. The controversial default setting Oh, the default setting caused a massive uproar? Because the release still goes live on its scheduled date, – unless the artist actively steps in and stops it. Right. Even though there is an advanced setting to require approval, it's not the default. Exactly. It's like a bouncer a club letting anyone wearing a VIP name tag into the VIP lounge up to the actual VIP to spot them and kick them out. That is a perfect analogy. You are basically making the artist act as a full time security guard. And the industry sentiment reflects that frustration. You had Sam Duboff from Spotify noting it was a highly-requested feature due to, quote, spammy releases or AI slop. on X at Spotify artists announced it and users like at Andy Beach pushed back hard. Arguing that approval should absolutely be the default. Right and at mid-er music echoed those concerns too. It's worth noting how the Synaptic Plunge team handles this. They Actually Utilizeunique, Unique Artists names claim their profiles early and turn all alerts on to prevent these mix ups. Which is smart digital high G. But looking at the legal side spot if I really slow case by case legal take down process into a quick dashboard switch. Yeah, because normally dealing with name likeness and false endorsement loss takes months and involves expensive lawyers. And Spotify is just trying to put a turn to get it on the problem right now. Exactly. So platforms are giving artists tools to protect their labels. But what do you the listener actually want when you press play? That is the multi-billion dollar question, honestly. And we finally have the answer thanks to Global Survey from Deezer and Ipsos. They polled 9,000 adults across eight different countries. The scale of that polling is really impressive. It is. And the staggering headline from this is that 80% of people want fully AI generated and AI assisted music clearly labeled. 80% is a massive consensus. It really is. And France actually leads that demand at 88%. And Deezer is acting on this data, right? Yep, they are rolling out a feature in December that will allow users to filter out fully AI generated tracks from of the search results entirely. Though it's unconfirmed, which platforms will get it first. Which is huge considering Deezer already bans fully AI tracks from their recommendations and charts. But they are fighting a tidal wave. Deezer receives 90,000 fully AI tracks daily. That accounts for over half of all daily uploads to the platform. This raises an important question about the perception gap among among listeners. What do you mean by perception gap? New October subscribers actually brought over libraries that had loss of creativity. So, start contrast to Brazil, though, where 61% still see it as a valuable tool., and And the reactions on. And the reactions on X were very telling at Deezer, news, confirm the 80% stat. And at vocal AI suggested that upfront labeling is just the smart move for creators now. Right. And at her four and three and three had that great quote label it, then battle it in the open. I love that. But it makes you think. Imagine discovering your favorite emotionally resonant track was made entirely by a server. It feels like a betrayal of trust. Yeah. The Deezer CEO noted that listeners feel much less attachment to AI artists because the human story is still the ultimate sticky label. It's the emotional anchor of the music. Exactly. So we've talked about labeling the output, like Suno visuals and filtering the output with Deezer. The biggest fight on the planet right now is about labeling the input. DESCENDANT Oh, this is the legal black box. What exactly are these models learning from? DESCENDANT Right. And before we get into this, we have to put a strict impartiality note here. These are all allegations and ongoing legal filings. DESCENDANT Yes. Nothing has proven no one is suing individual users and the judge hasn't ruled on any of this yet. TESTIE Absolutely. So detailing the US Corp. battle. Universal music group and Sony music are actively opposing Soonos request to pause a second lawsuit which the industry calls Soono 2. And that 2nd lawsuit involves 60,202 specific recordings. Huge number. So Soono wants to pause this second case until the Judge F. Dennis Saylor IV rules on the concept of fair use in the original Soono 1 lawsuit. Which makes sense from Soono's perspective. If fair use covers-covers them in case one, case two basically disappears. Right. But the labels fiercely oppose the pause because they want discovery, which over time, especially since a key engineer recently left the company. Pausing could freeze any real scrutiny until 2028. They want that data exposed to the light of day immediately. And this isn't just in the US, down in Australia, there's a massive front opening up. The Not AIs to take campaign launched on November 17 as well. It's a huge coalition. You have ARIA, APRM COS, independent labels, and legendary artists like Paul Kelly, Jimmy opt-out rule for copyright, and they are demanding a permission-first approach instead. So what does this all mean? It's a fundamental rewiring of digital ownership. Think of it with this analogy. An opt-out system is like a library allowing anyone to photocopy your deeply personal diary unless you proactively run down there and put a physical lock on it yourself. That is exactly what it feels like for these artists. The burden of protection is entirely on them. Right. And the stakes are just astronomical. If we unpack the stakes in the U.S. alone, statutory damages can reach $150,000 per work for, willful infringement. So if you multiply that by the 60,000 tracks. You hit a $9 billion theoretical ceiling just for this one case. Wow. And if the court rules it's not fair use, the entire economic foundation of these generative models changes overnight. And you see this anxiety everywhere. at music business and at I FBI..org and at news tongue X were all amplifying these legal stakes because it really is an existential Which require the same things from platforms like Suno. Right stems are the separated vocal drum, bass and instrument tracks. So you export those onto a Windows or Mac computer using DAWs like professional audio workstations. Right. You can't really do this properly on a phone. No, definitely not. But when you do that, you hit a massive problem called masking. Masking is such a headache for audio engineers. It really is. It's when AI stems fight each other. Like the bass bass, like the bass and the kick drum blurring together or the vocals getting buried completely under the synth pads. And to fix that, human engineers it's using some, like even tied split EQ 2. Which also had a huge update on November 17th with a $99 intro price. Naturally. Everything happened that day. I know right. So split EQ 2 is amazing because it separates the transient, which is the quick attack or the click of a sound, from the tonal part, which is the sustained body or the boom of the sound. It literally splits the physics of the audio wave in half. Yes. And it features dynamic bands and unmasking tools to make the different tracks actually talk to each other. And there's Oh, yeah. That's a two band dynamics plugin that features something called uplift, which which is upward compression, right? Exactly. It raises the quiet details like the tiny reverb tails or a singer's breath instead of just squashing the loud parts. Plus it has tube drive for that warm old valve amp saturation. If we connect this to the bigger this explains why human intervention still matters so much. Because AI audio is kon perfect. Basically, yeah., out of the box AI audio often lacks dimension. It sounds flat. Right. So by cutting the tonal part of an instrument where it collides with a vocal rather than just blindly turning the vocal up, the human engineer breathes life and clarity into the machine's these hybrid workflows. And again, the Synaptic Plunge team uses this exact cleanup pass before uploading their their Suno albums to YouTube. They don't,t just post raw AI audio. Never. But I want to bring it back to you, the listenern You don need to go out and buy a $100 plug-in to start doing this. No. , standard EQs inside any basic software can do this too. Exactly. It's all about the method. The actual magic happens when the human ear makes the subjective choice about what gets to take center stage. It's the human intentionality that makes it art. I Well said. We've covered a massive amount of ground today. Let's briefly recap our five main takeaways from this huge November 17 data drop. First, we had Suno visuals launching and embedding watermarks directly into their video files. Then we looked at Spotify's artist profile protection, trying to combat AI impersonation on the distribution side. The third was the title question directly. Yes, AI music should absolutely come with a label. The data and the industry trends all point to yes. They really do. But I want to leave you with one final, slightly provocative thought to mall over. As AI models get better and better at mimicking flawless pristine audio, human creators are going to increasingly rely on their intentional floss. That's an interesting point. Right. Like the raw breath before a vocal line, the slight misalignment of a drum. The analog warmth and chaos of tube drive distortion. They will use those flaws to prove their humanity. The ultimate label of authenticity in the future might simply be imperfection. So I encourage you to look at your own Spotify or Deezer playlist today with fresh eyes and entirely fresh ears.