I believe I can fly, sung by Michael Jackson? Wonderwall by Kurt Cobain? You got it.
As someone working in the industry… I better start getting better at my woodworking skills. Damn.
One thing is that the output of AI is not copyrightable. It can create all kinds of stuff, but it’s all public domain.
How does this work? I’m listening to a “Draik” song and I’m not familiar enough with drake to know the difference. The song sounds like the original. What is different?
I don’t know any Dake song, but at least for other “Artists” they only have songs that those artists never did. Cover versions. Or in Michael Jacksons case, Jackson5 Michael “singing” his later stuff.
So, sure the song you’re listening to is a Drake song?
This is going to do a lot of damage to smaller music-makers unfortunately. This is going to become pretty much all we will get from now on for background music applications.
That’s already a reality. I work in post-production, and tools like https://soundful.com/ are starting to replace “classical” stock libraries with their vast but limited content. We have already seen first clients switch to AI voice over instead of voice actors, too.
While this is technically impressive, these are still derivative works. The copyright holders of the song still would need to give permission for performance and still would get paid when they are performed or used commercially. Thriller sung by Freddie Mercury? Neat. But I can hear a Michael Jackson cover band live that could do the same thing. There is nothing novel about song covers we just happen to know who is not known for a particular song.
I’m the 50s and 60s artists would cover songs within a year of the original release. Elvis had a bigger hit with Hound Dog, but it wasn’t the first version of it. For the moment, AI can’t tour, have an image, get caught in a compromising position, etc that creates fan bases. I’m not worried about this really taking away any jobs yet. Everyone knows these songs - let’s talk when AI charts with an original song that is not based upon some one else’s style.
I don’t think this site is interesting because of the creativity. It is - as you said - not there. But the voice models are next level compared to what we have heard before.
How so? Vocaloid already exist for nearly a decade, back then many are worried that it will replace real human singers (because why hire singers when music composer can just use their computer?)
Fast forward today, pretty much only Japanese that still use it. Even during the height of it’s popularity, it didn’t even manage to “fully replace” human singer.
Vocaloid isn’t ai. Its a voice synthesised from a real singers voice. The characters attributed to the voices are just ai in lore
Really impressive
