Independent Label CEO on Music AI and Future of the Industry

AI-generated music is evolving fast, but is it ready to replace human creativity? We hear from an independent label CEO on music AI. He tested 13 music AI tools and shares insights on the future of the industry, the risks for artists, and why human-driven music may become more valuable than ever.

HYPEBOT EXCLUSIVE

Independent Label CEO on Music AI and Future of the Industry

by Mark Shishlev from independent Label Cantroll

Years ago, I started my own music label. It’s underground. It’s a niche label, but the residents include artists from every continent in the world. And now with the widespread adoption of neural networks, I’m thinking: What’s next? And how do you feel about the fact that it used to take a week or a month to write a track, and now it takes a minute.

Hi. This is Mark, founder and head of Label Cantroll. I’ve tried 13 music neural networks (Suno, Musicfy, Musichero, Elevenlabs, Boomy, Udio, Melobytes, Soundful, Beatoven, Soundraw, Soundverse, Amper and Stable Audio). Gathered my thoughts on the new industry in this post. 

Present

Right now the AI output is very sparse: groovy, but not serious. With big problems in terms of sound quality, especially for beatless versions and experimental genres. Most often, services give results with bitrate of only 192 kbps (Suno, Udio, Musichero) or 127 kbps (Stable Audio). In all of the above services on paid accounts the results of generation can be downloaded in WAV format, there is a function (or planned development) to divide tracks into stems. 

Some services have their own features on free accounts: 

• Musicfy generates music segments of 15 seconds or less;
• Soundverse up to 30 seconds;
• Beatoven up to 60 seconds;
• Elevenlabs up to 22 seconds; 
• Soundraw, Boomy and Soundful turned out to be generators offering to make a combination of suggested tracks with suggested settings. 

I believe that these services are not suitable for use in a professional environment of musicians, producers or customers – movie companies and TV channels. Not only because of the sound quality, but also because – the real meaning in creativity comes only from a human being. It is important to understand: neural networks have no life experience, no intelligence, it does not think or feel. AI models used to be simple combinatorial generators. Now the algorithms are much more complex, but AI still uses common and obvious solutions. For example, if a promt contains the word “beats”, then surely the melody will contain electric piano or keys. And if “lofi” – then it will contain light noises in the background. However, most of the above services are not meaningless and can be positioned as an assistant for musicians and beatmakers – for example, to find a sample.  Also, can be useful for ordinary users for household use or for a joke. For example, for the background in stories or to “record” a song for someone’s birthday. Technology is evolving rapidly. Everything will change in a year. 

I keep five more services in mind – at the moment, they are either not in open release or are available by invitation only. These platforms are developed by corporations – MusicLM from Google, GenAI from Adobe, MusicGen from Meta and Fugatto from Nvidia. Also in my bookmarks is Ripple by ByteDance, an app available in the App Store only in the US.

Future

Not only venture capital startups but also corporations invest billions of dollars every year in the development of neural networks. They expect that the quality of their product will grow and therefore in a couple of years there will be no differences between neuromusic and instrumental music.

Musicians learn by listening to other musicians. Artists did too. Already on the basis of this, having reworked and been inspired, they wrote something of their own. In general, this is how human learning happens. The brain generates something from what it has seen, heard or felt earlier. Neural networks work in exactly the same way – they generate based on someone else’s experience. This is the position that developers of AI services will hold in courts against Sony Music and Warner corporations on the topic that neural networks were “fed” with illegal copyrighted files. But on the legal side, everything will be clean.

Streaming services pay artists little (a separate topic!), and if they want to pay even less, they will include neuromusic in their playlists. Perhaps they will generate this music themselves. The logic is simple: the main expenses of streaming platforms are royalties. And in order to reduce costs, algorithms should be able to feed listeners something they don’t have to pay for. If not already, it won’t be long before they hire employees or an agency to give them thousands of neuro tracks for a salary. You will see hundreds of obscure like-for-like real musicians with millions of auditions. Such “artists” with the same and universal music and songs will get their own place in playlists and individual streams (like “My Wave”). It’s a controlled model of content consumption. I also believe that only smaller platforms will not take advantage of this. After all, to increase subscription sales, they need massive posts from labels and artists enticing you to listen to new releases from them. Already, Deezer has announced that they will exclude AI tracks from recommendation systems and editorial playlists. 

The use of generative technologies will become commonplace. It will be possible to generate beats and songs for yourself in your smartphone. I hope saturation comes quickly, but I’m sure we’ll be seeing headlines like:

• AI band performs with video avatars
•  a lifestyle blogger bot, who has all the photos, text and videos generated.
• stocks platforms with All-AI photo and audio content 
• indie game or TV series with AI soundtrack
• albums of deceased rappers or singers with all AI: music, lyrics and a clone of the voice in all vocal parts.   

What are we preparing for?

To be even more creative. We need to sing and play in a way that’s different from the robot. Or rather, from what we’ve uploaded to the neural network as an example. Strive to become not just a band or singer, but a charismatic personality with active social media accounts and concerts. Because, this creates added value to your project.

Perhaps the creativity made by people will become a competitive advantage and something more exclusive. Someone will add to the descriptions of music albums that “lyrics and music are written without the use of AI”. But … if you want to make money from music – it’s important to realize – the competition is huge! Singers, beatmakers and musicians are not required as much now. You can’t be banned from creating and playing guitar with the guys at the rehearsal base. But you can neuro-fake devalue the very practice of playing instruments, beatmaking, or painstakingly mousing in programs. You have to be willing to not be disappointed. Good luck with everything!

Collected a retrospective playlist, with the best tracks released on Label Cantroll from 2023-2024.YouTube Music | Spotify Made by humans!