German company Musical Bits https://musicalbits.de/ has released the first single of their AI virtual heavy metal band "Frostbite Orckings". https://www.orckings.org/?view=article&id=37&catid=8.
Musical Bits creates software that creates music, with the support of AI. Our Maisterstück platform uses AI technology to model all layers of creativity of a human composer and implements these layers as reusable and combinable software components. Maisterstück's functionality can be accessed via a service oriented API.
The Musical Bits software can create music from real time data, from various user interfaces or from our own emotion modelling engine EME. We even create full virtual bands, albums and songs. For example, check out the Frostbite Orckings.
Their sound could be described as a keyboard-heavy version of viking metal, like a mellower spin of Amon Amarth. Along with the song comes an also AI generated video https://www.youtube.com/watch?v=2EaJCt2GpVc of orcs playing along. At this time, it is unclear what input has gone into the AI to generate the production, and how much post processing is done.
The path seems to be set into a direction where we simply can run text-to-song AI ("AI, play me a new Motorhead song with lyrics about whiskey") in the foreseeable future and get convincing results.
(Score: 0) by Anonymous Coward on Tuesday March 21, @09:43AM (2 children)
https://www.youtube.com/watch?v=Emidxpkyk6o [youtube.com]
https://www.youtube.com/watch?v=4Q1JBwC0SqY&list=PLG89Sukk5_MvY0HdwmcED75OOx215iu4s&index=1 [youtube.com]
(Score: 5, Informative) by driverless on Tuesday March 21, @10:01AM (1 child)
You forgot the original(?) AI-generated metal, Dadabots Relentless Doppelganger [youtube.com] which has been running for something like four years now. Now that's real metal, generated by real... umm... real...
(Score: 4, Interesting) by Rich on Tuesday March 21, @04:53PM
TFA Submitter here. IIRC, RD is trained on audio samples alone, which doesn't allow for variety much. It's a bit like the current (or at least last month's) state of StableDiffusion for images. Nice for some creative ideas if you stare at a white sheet, or some artistic one-off, but nothing that can be the base of an entire virtual world.
They are rather quiet about the steps they go for a production. Do they go straight to audio and adjust that right, or do they have an intermediate representation? E.g. "Create a bass line in the style of (Geddy Lee | Steve Harris | Lemmy)" that goes to MIDI first. And then an instrument render? "Play the bass like " that would provide the audio at instrument out. Next a backline render: "Re-amp the raw bass audio through the Murder One amplifier stack, picked up with an SM57 in center position". And finally "Give me a chart-breaking Rick Rubin style mix?". (A future end-user word-to-song release could then have another layer that translate the layman request into the required setup of sub-components and instructions for those.
Their job page has an opening for producers, so intermediate steps are likely involved. Also, they make the claim that they had "world class" musicians recording the training data. They clearly want to dodge the coming lawsuit wave from old media. I'm not sure what to think about this. The released song definitely sounds like the AI has heard some Amon Amarth during training.
(Score: 4, Interesting) by Rosco P. Coltrane on Tuesday March 21, @09:48AM (2 children)
I actually have high hopes that it will improve a lot today's "music", because the bar is so low.
And before anyone accuses me of being an old fart going "kids these days", I'll point out that a lot of music has been utter crap for many decades and could benefit from the removal of the "musicians" (sorry I meant "content creator") from the loop. I guess an early example of that was Milli Vanilli, and if I recall correctly, it was quite successful, faked though it was.
(Score: 1, Informative) by Anonymous Coward on Tuesday March 21, @12:28PM
"Hey hey we're the Monkeys . . ."
One problem is music means different things to different people and they have different expectations. Things like K-pop are basically the same thing as Milli Vanilli and hugely popular. Some people get huffy about people not writing their own music, but on the whole, very few people do. Commercial music has always been targeted for mass appeal and if any particular genre pops out of that, the music companies quickly get out ahead of it and commercialize and control it.
(Score: 2) by takyon on Tuesday March 21, @12:38PM
https://www.youtube.com/watch?v=GOw7jHLij2A [youtube.com]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by fliptop on Tuesday March 21, @10:41AM (2 children)
I like that the bass player is left handed. Well done.
To be oneself, and unafraid whether right or wrong, is more admirable than the easy cowardice of surrender to conformity
(Score: 2) by Rich on Tuesday March 21, @04:30PM (1 child)
Mio Akiyama fan, then?
I liked how the little drummer orc rode on the back of the big bass drummer orc with his kit. Let's hope that, because they're so cute, the elves relent and don't just use them for target practice. ;)
(Score: 2) by fliptop on Tuesday March 21, @06:52PM
Actually, I was thinking more Paul McCartney.
To be oneself, and unafraid whether right or wrong, is more admirable than the easy cowardice of surrender to conformity
(Score: 2) by looorg on Tuesday March 21, @11:17AM (1 child)
Not bad, I have heard worse music of the genre. Still it does sound a bit like metal-elevator-muzak. Which isn't really a good aspect.
(Score: 3, Insightful) by JoeMerchant on Tuesday March 21, @12:16PM
>sound a bit like metal-elevator-muzak. Which isn't really a good aspect.
I'd say that depends on the audience, in a Finnish department store catering to the 50-70 year old clientele it might just be the ticket.
Of course this is early days for the technology, I foresee audience feedback driving the algorithm. Facial analysis, other physical/emotional response measures, driving the composition and performance style to desired effects. Couple this with virtual audiences around the world and the AI might play to a virtual audience of 100,000 but divide that audience into 6 or 8 distinct populations playing "to the tastes" of each population, and maybe reshuffling the listeners/viewers among the various performances to find the best fit for their tastes.
Of course, you should expect laser focused advertising styles to follow you around after baring your soul to such an event...
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 2) by DannyB on Tuesday March 21, @02:41PM (3 children)
Maybe a musically inclined AI should generate all possible musical phrases, of gradually increasing lengths.
Then this could be used to either:
1. defend against ridiculous copyright attacks that your own original music is copied from some tiny musical phrase somewhere
2. sue (or countersue) that someone's music is an infringement on the AI created music, for various definitions of music.
How often should I have my memory checked? I used to know but...
(Score: 4, Insightful) by JoeMerchant on Tuesday March 21, @03:35PM (2 children)
Maybe AI generated music should put an end to music copyright altogether.
The thing that's semi-broken in both copyright and patent IP protection is the protection of minimal investments. Submarine patent trolls who invest $20K (or less) each into 1000+ patents, then wait for the $20M+ payoff to appear in one of the thousand, sue, profit, rinse, lather, repeat.
If AI can generate songs for pennies in electricity, those songs shouldn't be copyrightable until they have had at least as much invested in them as it would have taken to publish a song in 1900, adjusted for inflation, cost of labor, etc. Preferably more.
I'm not suggesting that you can't take an AI generated song, promote into the public consciousness, and then enjoy copyright protection on that investment. I am suggesting that an AI might publish all possible melodies (Melancholy Elephants style) into a "publicly available website" and then troll the airwaves and digital channels looking for matches to it's "previously published works" then sic an AI lawyer on the person(s) who promoted the song seeking damages. And I sincerely believe there are better uses for our legal system than that.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 0) by Anonymous Coward on Tuesday March 21, @08:24PM (1 child)
There should be no implicit copyright. You want copyright, you submit a copy to the national library in each country and pay a yearly registration fee. No fee and it hits the public domain a year later.
I would consider it reasonable for countries to have agreements to respect each other's copyright, but only if a search in one national library automatically brings up results in all such libraries.
(Score: 2) by JoeMerchant on Wednesday March 22, @12:25AM
That is a very sensible system. And therefore highly unlikely to be implemented.
Україна досі не є частиною Росії Слава Україні🌻 https://news.stanford.edu/2023/02/17/will-russia-ukraine-war-end
(Score: 2) by richtopia on Tuesday March 21, @07:37PM
If you aren't familiar, Muzak is a brand of background music. The generic music you'll hear in factories with no vocals. I stream on Twitch, and I would like music that is royalty free, never ending, and blends into the background. Bonus points if the AI can watch the stream and change tempo accordingly.
(Score: 2) by oumuamua on Wednesday March 22, @03:08AM
in the style of symphonic metal with lyrics of a metal band's thoughts of being replace by AI https://www.youtube.com/watch?v=P7vg7ir8jyc [youtube.com]