Page 4 of 5 FirstFirst 12345 LastLast
Results 76 to 100 of 113

Thread: YES ... or not

  1. #76
    Mod or rocker? Mocker. Frumious B's Avatar
    Join Date
    Nov 2012
    Location
    Smyrna, GA
    Posts
    1,160
    There are prog records by real bands that are just as incoherent and lame so why is that proof of the music being created by AI?
    "It was a cruel song, but fair."-Roger Waters

  2. #77
    Member
    Join Date
    Nov 2012
    Location
    Wisconsin
    Posts
    754
    as many have said if you start listening to AI generated music you can recognize all sorts of digital artifacts, the likes of which are all over this thing. nearly all the AI-generated stuff I've heard sounds like a 96 kbps MP3. you might notice that none of the instruments are really all that well "defined". everything sounds sludgy. its like the entire band are playing in a room and someone's recording them on their phone 500 feet away. you can't really separate any of the elements out and if you try to focus on any of them you get a real uncanny valley feeling.

    I do find it kinda impressive though. it does sound like Yes, but the way it blends together different eras of the band is kinda interesting. that said it's not as if someone just went into Udio and prompted it to generate a Yes album, I believe making something like this requires a lot of specific prompting. it's not *good* though. it's like what would play in your head if you just tried to imagine a Yes epic on the spot. maybe someone could take the plunge and listen to this 3-4 times but it's kinda wild that there's not a single thing I can remember about it. like I can kinda remember the structure but the melodies...nope. even the Jon Davison Yes albums were way more memorable than this.
    Critter Jams "album of the week" blog: http://critterjams.wordpress.com

  3. #78
    Jazzbo manqué Mister Triscuits's Avatar
    Join Date
    Nov 2012
    Location
    Utopia
    Posts
    5,511
    Quote Originally Posted by JAMOOL View Post
    even the Jon Davison Yes albums were way more memorable than this.
    Well, let’s not go overboard.
    Hurtleturtled Out of Heaven - an electronic music composition, on CD and vinyl
    https://michaelpdawson.bandcamp.com
    http://www.waysidemusic.com/Music-Pr...MCD-spc-7.aspx

  4. #79
    Quote Originally Posted by Frumious B View Post
    There are prog records by real bands that are just as incoherent and lame so why is that proof of the music being created by AI?
    The proof is in the digital artefacts, the laughable genericness and indeed the incoherence. There is absolutely no mistaking it if you’ve heard some of these.

  5. #80
    cunning linguist 3LockBox's Avatar
    Join Date
    Nov 2012
    Location
    hiding out in treetops, shouting out rude names
    Posts
    3,708
    I want A.I. to do Starcastle or Cathedral or Druid next.

  6. #81
    Member Digital_Man's Avatar
    Join Date
    Nov 2012
    Location
    Philly burbs PA
    Posts
    5,754
    Quote Originally Posted by 3LockBox View Post
    I want A.I. to do Starcastle or Cathedral or Druid next.
    Why Cathedral?
    Do not suffer through the game of chance that plays....always doors to lock away your dreams (To Be Over)

  7. #82
    Kind of curious what an AI Gentle Giant or King Crimson would sound like.

  8. #83
    Member moecurlythanu's Avatar
    Join Date
    Nov 2012
    Location
    The Planet Lovetron
    Posts
    13,302
    Quote Originally Posted by pb2015 View Post
    Kind of curious what an AI Gentle Giant or King Crimson would sound like.
    They would sound like AI.

  9. #84
    Quote Originally Posted by pb2015 View Post
    Kind of curious what an AI Gentle Giant or King Crimson would sound like.
    Easy enough to find out! Here's a quick bash at some AI Gentle Giant using Udio. It's not great (to say the least), and a lot less convincing at aping a specific artist than the Yes example at the top of this thread, but it hopefully goes some way towards convincing the doubters that this technology is real and here to stay.

    Context, for those who care - I would consider myself to be passingly musically literate, but by no means a musician or composer. This piece was generated using a single text prompt "funky, complex 1970s progressive rock with counterpoint with lyrics about making a sandwich" (Udio doesn't allow you to specify artists so that was my attempt to succinctly summarise GG's sound circa Free Hand/TPaG/Interview and give it a suitably inane topic - in my defence it is currently lunchtime in the UK!). My only other creative input came from specifying which direction I wished the piece to be extended, and picking between two generated outcomes each time - all musical and lyrical content was 100% AI generated. The whole process took about half an hour, but most of that was taken up with waiting for the new clips to be generated.

    So thinking about this in the context of the Yes "album" that kicked off this discussion, I think we can probably conclude that the person responsible used separate software to generate the vocals, and also presumably generated hundreds, perhaps even thousands of clips to draw upon, rather than the ten I generated to "create" this example (of which five were used and the other five discarded). So a not insubstantial amount of effort, but little to no actual creativity or musical ability required.

  10. #85
    Member
    Join Date
    Nov 2012
    Location
    Wisconsin
    Posts
    754
    lmao, I hear just enough in there to think the model was trained on Gentle Giant, but it's kind of a mess otherwise

    I tried to get it to generate a "King Crimson" song, if you listen to the end of this 30 second clip you can tell it knows what I was trying to get at lol

    https://www.udio.com/songs/7rWKLhE8wLmPTZENN2L4mu

    whatever was used to generate the Yes album I suspect took a lot of specific prompting, as well as training on Jon's voice specifically. I know there's software where you can upload specific pieces of music and have it generate something like that. amusingly this sounds like it was trained more on like, Keys to Ascension-era Yes
    Critter Jams "album of the week" blog: http://critterjams.wordpress.com

  11. #86
    Quote Originally Posted by JAMOOL View Post
    lmao, I hear just enough in there to think the model was trained on Gentle Giant, but it's kind of a mess otherwise
    Oh yeah, it's bloody awful! But there's moments, perhaps more down to happenstance than anything approaching "intelligence" (artificial or otherwise), that could be charitably interpreted as Giant-isms. I think trying to create complex music really stretches the algorithm to breaking point, as it wants to keep everything in 4/4 with conventional chord progressions. If you listen to the examples from more formulaic genres like country or EDM it's far more convincing.

    The point is that I managed to churn that out with just half an hour of messing about on my lunchbreak. It's not hard to imagine that you could generate far more accurate results by dedicating a substantial amount of time to trying different prompts and wading through hundreds of options.

    Quote Originally Posted by JAMOOL View Post
    whatever was used to generate the Yes album I suspect took a lot of specific prompting, as well as training on Jon's voice specifically. I know there's software where you can upload specific pieces of music and have it generate something like that. amusingly this sounds like it was trained more on like, Keys to Ascension-era Yes
    You may well be right. As I say, I think it's highly likely the vocals at the very least, were created using a different system. Udio tends to make everything sound vaguely "modern" even when you specify earlier decades, so it's not suprising to me that digital-era Yes is the best it can manage.

  12. #87
    Member Plasmatopia's Avatar
    Join Date
    Nov 2012
    Location
    Plague Sanctuary, Vermont
    Posts
    2,546
    "Cheese so sharp, it cuts the night"

    <sig out of order>

  13. #88
    Quote Originally Posted by runciblemoon View Post
    Thanks. I like how at the beginning it sounds sort of like "Raconteur Troubadour" before eventually turning into more of a 2000's pop thing with a vocalist in Derek Shulman's range but sounding more like Sting.

  14. #89
    Quote Originally Posted by JAMOOL View Post
    amusingly this sounds like it was trained more on like, Keys to Ascension-era Yes
    That's what it sounds like to me. Have an A.I. mimic the studio tracks on the KtA albums.
    Mongrel dog soils actor's feet

  15. #90
    Member
    Join Date
    Nov 2012
    Location
    Iowa City IA
    Posts
    2,494
    Quote Originally Posted by JAMOOL View Post
    lmao, I hear just enough in there to think the model was trained on Gentle Giant, but it's kind of a mess otherwise

    I tried to get it to generate a "King Crimson" song, if you listen to the end of this 30 second clip you can tell it knows what I was trying to get at lol

    https://www.udio.com/songs/7rWKLhE8wLmPTZENN2L4mu
    Most of your clip seems to have been trained on Kobaia-era Magma! Its a catchy drum pattern. I think if composers used clips like this one to generate ideas rather than as final product the effect of AI might actually be a net positive.

  16. #91
    Quote Originally Posted by Plasmatopia View Post
    "Cheese so sharp, it cuts the night"

    That’s the line that got me too 🤣🤣🤣

  17. #92
    Whereas it might as easily be said, "Music so pathetic, it cuts the cheese."
    Cobra handling and cocaine use are a bad mix.

  18. #93
    I wonder how crazed people will get when A.I. music is accepted by the masses and no matter how much people scream and yell about art and integrity and this or that, the audience simply won't care.
    Mongrel dog soils actor's feet

  19. #94
    Member
    Join Date
    Feb 2023
    Location
    Southern California, US
    Posts
    505
    Quote Originally Posted by Splicer View Post
    I wonder how crazed people will get when A.I. music is accepted by the masses and no matter how much people scream and yell about art and integrity and this or that, the audience simply won't care.
    Rick Beato will undoubtedly have something to say about it. According to him the state of current popular music has already slipped down a technology-driven slope; this will merely accelerate the existing trends.
    What we feel we have to solve is why the dregs have not dissolved.

  20. #95
    Quote Originally Posted by Splicer View Post
    I wonder how crazed people will get when A.I. music is accepted by the masses and no matter how much people scream and yell about art and integrity and this or that, the audience simply won't care.
    "No one ever went broke underestimating the intelligence of the American people." -- H.L. Mencken
    Cobra handling and cocaine use are a bad mix.

  21. #96
    Quote Originally Posted by Sturgeon's Lawyer View Post
    "No one ever went broke underestimating the intelligence of the American people." -- H.L. Mencken
    People have made a lot of money not underestimating the desires of people to want what they want when they want it. -- Me
    Mongrel dog soils actor's feet

  22. #97
    Quote Originally Posted by runciblemoon View Post
    Easy enough to find out! Here's a quick bash at some AI Gentle Giant using Udio. It's not great (to say the least), and a lot less convincing at aping a specific artist than the Yes example at the top of this thread, but it hopefully goes some way towards convincing the doubters that this technology is real and here to stay.

    Context, for those who care - I would consider myself to be passingly musically literate, but by no means a musician or composer. This piece was generated using a single text prompt "funky, complex 1970s progressive rock with counterpoint with lyrics about making a sandwich" (Udio doesn't allow you to specify artists so that was my attempt to succinctly summarise GG's sound circa Free Hand/TPaG/Interview and give it a suitably inane topic - in my defence it is currently lunchtime in the UK!). My only other creative input came from specifying which direction I wished the piece to be extended, and picking between two generated outcomes each time - all musical and lyrical content was 100% AI generated. The whole process took about half an hour, but most of that was taken up with waiting for the new clips to be generated.

    So thinking about this in the context of the Yes "album" that kicked off this discussion, I think we can probably conclude that the person responsible used separate software to generate the vocals, and also presumably generated hundreds, perhaps even thousands of clips to draw upon, rather than the ten I generated to "create" this example (of which five were used and the other five discarded). So a not insubstantial amount of effort, but little to no actual creativity or musical ability required.
    Thanks. That's interesting to hear and read about the process you went through. It is impressive. I mean, it's not great as music, but it's impressive what the AI can produce, and the AI will get better. The question becomes what will the AI plateau at, and what are the use cases for the result. You can see incidental music in TV and film going this way.
    Where Are They Now? Yes news: http://www.bondegezou.co.uk/wh_now.htm
    Blogdegezou, the accompanying blog: http://bondegezou.blogspot.com/

  23. #98
    Quote Originally Posted by Plasmatopia View Post
    "Cheese so sharp, it cuts the night"

    THAT has to be JA.

  24. #99
    Member
    Join Date
    Nov 2012
    Location
    Iowa City IA
    Posts
    2,494
    Quote Originally Posted by Drake View Post
    THAT has to be JA.
    I would say more like Sinfield on a particularly good day.

  25. #100
    Member
    Join Date
    Feb 2023
    Location
    Southern California, US
    Posts
    505
    Quote Originally Posted by arturs View Post
    I would say more like Sinfield on a particularly good day.
    It's at least as good as "Bareback ladies have fish."
    What we feel we have to solve is why the dregs have not dissolved.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •