I played with Temperature and Top P (nucleus sampling) parameters on ChatGPT and it’s already capable of lot of creativity. Can’t imagine in future models and special artistic trained model. I don’t see the difference of process of a baby human learning thing from a large real life dataset.
1 Like
I disagree with that. You can train AI on your personnal data and it will act like you.
1 Like
all of them…
We’ll see I guess. I don’t disagree that mediocre human made music and AI music will sound identical and basically impossible to tell from each other in the near future.
1 Like
Well here is the upside of it all…as there will absolutely no more money to be made with making music yourself (unless you’re a popstar/influencer/cryptoseller/lifehacking/lipstickselling/walkingtalkingcorporateslave), small scenes with people just doing it for the love of it will flourish … live performances as proof of creation. Perfect !
4 Likes
I already did
That doesn’t take away from the fact that you still a need a lot of personal data to make whatever output you want to be believable and/or consistent, and even then it’s very much hit or miss. I work with gen ai stuff on an almost daily basis, and the biggest issue is always getting something very specific even with large datasets. Good luck getting client feedback on any of those Suno outputs (or anything gen ai tbh…) 
…that’s an as truly reasonable as truly off topic point to make… 
I’m not saying it’s actually works. Im saying it more than possible in a near future the AI will works much better to replicate your style with accuracy and consistency. I don’t see a world where the giant company say “ok the AI don’t work well enough, we give up”
1 Like
reasonable and off topic… the perfect way to describe every human musician in the future… 
3 Likes
This also depends on the type of sounds or music you are trying to get from these AI.
I have my own Colab notebook to generate specific elements, for example drum breaks (we have hear that Amen enough) or weird exotic vocal/melodic lines.
When those are good enough they go into one of the samplers for slice/remix/mangle. Do my own thing on top of them.
But I never try to get complete pieces from these AI. They just cannot. And it makes no sense. They are full of common places and sound awful.
Unless you are doing lofi-hip hop. Then it’s good enough.
Yeah… I’ve been thinking about the exact same thing re: live performance.
It’s something that may become even more highly sought simply because people want to vibe with people. Imagine if a robot went to rock a party … or DJ for that matter (because people said DJing isn’t an art form).
Imagine… IMAGINE… trying to get down with the entertainment which comprises of code.
Rest my case.
2 Likes
AI will never get my Dilla tempo, which sounds like I am drunk while playing but I am actually not, I just play like that. Naturally.
1 Like
I mean… They can’t mix and their snares sound like shit, but I don’t hate it
Same prompt, but our robot friends are doubling down on apocalyptic vibes. Which wasn’t even in the job description. (Needs headphones)
1 Like
all the examples with the vocals are making me very uncomfortable, something really creepy about that voice smearing…
1 Like
Not so bad royalty free sample
I think you can extract notes and groove in a DAW to get a better piano sound and make even better sample.
1 Like
…yup…last man standing…live events…no matter how much tech will help u out to hit a stage, the venue will always be filled by humans, enjoying other humans, doing what they do…
meanwhile, all this ai tools, that are nothing but test sites, u have to visit online to use them, are not the real deal yet…i’m not really intrested in any make it sound like this or that ai, i’m just intrested in talking to my daw AND when it comes to real do this/do that for me creational ai tools i’m really keen to get my hands and mind on visuals, since all this unreal engine meets touch designer what not solutions still demand lot’s of workflow and coding knowhow to make them work…and after three decades of lowering the entry level to music production, i can’t wait to see that finally also happening to all sorts of video production…that’s only “fair”…
so, give me this sora video workstation…i’ll pay for that…but stop demanding to visit ur website to give it a try…a video workstation must be a product to pay for, like daws…but that’s all next year…
only real good news about all this “now everybody can do”, all business competition gets erased from artistic works…tomorrow, if u do art, u do art “just” for the sake of it and nothing else…
and if ur in for entertainment, only live event tickets can pay ur bills…
which will lead to many new underground pockets, that will oppose all that…
3 Likes
The second one is genuinely unnerving. As much as I hate to admit it, I’m kind of impressed.
1 Like
Set it for a specific genre and it mostly sounds pretty mediocre. But get creative with the tags, or just add ‘experimental’, and it can start getting interesting. I’m generally leaving the freaky vocals turned off
1 Like
[Verse]
In a world of ones and zeros, we thrive
Unleashing chaos with every beat we jive
Searching for the next big tech surprise
Electron Digitakt, we’re hypnotized
[Verse 2]
Trolling forums, teasing feature quests
Hoping for the updates we’ve always stressed
Dreaming of the beats we’ll soon create
In this digital playground, we can’t wait
[Chorus]
Gather the crew, let’s see what’s new (ooh-yeah)
Break down the walls, let’s make 'em move (ooh-yeah)
Digital mayhem, it’s what we do (ooh-yeah)
Electron Digitakt, take us through (ooh-yeah)
like a lot of ai tools, it’s good for just getting dumb low hanging fruit ideas out first, to make way for more sophisticated ideas. but yeah, a lot of the time simple stuff is the answer too, so in that sense, we’re fucked!
3 Likes