I’ve spent some time with @suno_ai_ today, and my conclusion is that this #AI tech is going to be more disruptive than #napster was to the music industry.
As an example, I attempted to create a Perfume – style Japanese electro song about chocolate and dancing – only described the type of feel, lyrical theme and genre and got an output in about 20 seconds that tracks well towards what I wanted in terms of output.
Amazing.
Take a listen for yourself. https://suno.com/song/23bbb498-cb03-4fb6-8d7d-44daaba9a7c3
Personally, while I don’t believe this tech is going to replace really great well-crafted real music, and its not at that level yet, it’s going to completely destroy the utility of music libraries.
For example, if I’m a creative director of a laundry detergent brand at an ad agency, why would I go to a music library when I could, as I did, input that I wanted an afrobeat song extolling the virtues of laundry pods and then get this thrown at me in 20 seconds:
https://suno.com/song/19481a6e-2dca-4817-8553-5fe2e107419e
It needs work, but the fact it can be done at all is both miraculous and frightening. That said, there’s a lot about it that gives me pause.
I made a few requests using band names. That seems like a red line I shouldn’t be able to cross, but 2/3 of the searches allowed it.
First, I asked for a song about Hades that sounds like Sisters of Mercy.
It’s definintely not at the SOM level, but the fact it output this seems like its screaming for a lawsuit:
https://suno.com/song/00b182da-f29b-499e-a210-19b3caf6cc2e
And then, going back to Perfume – I then put in a search asking for a song about rabbits that sounds like the group Perfume, and I got this:
https://suno.com/song/ce9234fc-e69a-485b-8101-adfdf219f99e
Again, screaming for a lawsuit. The algo even used the band’s name in the lyrics. But waIt… when I asked for a song about Peeps that sounds like Rammstein, it did stop me, saying – “Couldn’t generate that.
Song Description contained artist name: rammstein”
Hmmm… ok – well, I guess they have some kind of firewall against copyright infringement for really well known bands. How about if I ask it to create a neue deutsche härt song (Rammstein’s genre) about aspirin…
https://suno.com/song/5e72733c-f4c4-46e2-a182-d513e24accb0
Or this neue deutsche härt song about Hamsters, which, without prompting, the AI took to a very dark place lyrically:
https://suno.com/song/738922f5-f70b-4bc2-94e4-dabe522a6bfa
Well, neither is Rammstein level, but if you’re a creative director at an ad agency it may be good enough. Or at least a jump-off point for further refinement with a live band.
Second, it’s obvious from my research that the algo is being taught off of actual music, and there’s going to have to be a reckoning and a realignment of their business model because the rights owners are going to come a calling sooner rather than later, torts in hand asking for a share of whatever action Suno is getting – and rightly so.
But, the music business being what it is, and the tech business being what it is, Suno and companies like them are going to launch and build out, then deal with the rights issues later if they reach critical mass – just like ringtone companies did in the early 2000s – go fast and break things, right?
That said, in sync and music creation for media, ads and brands, we’re a risk adverse bunch, and as I said, this is screaming copyright infringement, which will lead most music licensees to steer clear of placing this music into their projects for the moment – even if there are some AI-music based ad campaigns out there like Red Lobster’s Cheddar Bay Biscuits –
But these rights issues will be worked out at some point. I personally think it’s inevitable that AI-based music will be a portion of the music used for media. And as for creators? And the rights owners in general? What impact does all this have on songwriters and artists? Obviously, it does have some impact – the world of music libraries and music houses is going to evolve, and while some people will prefer to use only actual music from actually musicians, others will completely embrace this new tech.
I feel like there are two things that need to be done. First, there needs to be an advocation effort – a lobbying effort to Congress on behalf of the music industry to create some sort of pool of money to remunerate the people who’s work taught the algo.
Second, there needs to be a service – or a tier of an existing service where they bring in musicians and composers to teach the algo, and allow rights holders to get paid a fee for premium melodies, beats and rhythms that have been fine-tuned and/or created by real artists – so a blending of original works and AI.
I’m going to have to do a good deal more thinking about this, and I’m definintely going to write more about this and its ongoing evolution and implications for all of us, but I’ll end with saying this is the only time since I downloaded Napster for the first time 25 years ago that I said to myself “wow, this tech is amazing and it’s going to totally destroy the industry as we know it.” Let’s hope I’m wrong this time and the industry actually gets it’s act together before it’s steamrolled over again by tech.