Latest on MJJC

A.I Michael - opinions

wendijane said:
I assume Michael would be angry...i think hed be outraged you know.
Click to expand...
I think most of us here can agree, regardless of our feelings on AI, or the ethics, that MJ would love this. At least be curious about it. Anger would be the absolute last thing on his mind.

Would he outright cosign it, like Grimes did? Maybe not. But after a while he might experiment with it. He had his own process for making music anyway.

I don't want an AI future of MJ music either, I'd rather hear the few complete tunes and then the demoes in a box set, then we can probably stop. Don't ever create AI music. And probably make duets out of the incomplete stuff, or tribute covers.
 
Mister_Jay_Tee said:
That's what I thought until I heard the Hello Twitter thing. Now I say they'll crack the code sooner or later.
Click to expand...
This point is the worst point AI will ever be at. Every time you hear a truly competent recreation, already, just imagine in 5 years what it can do. The misinformation can be insane.
 

I don't know if the songs on the channel really are leaked demos made for MJ but this, to me, is the best one. Would love to hear an A.I cover of this.
 
im so convinced he would be dead agasint AI




a childhood dream here.... - I am sorry Teddy :cry: ..sorry bobby,.... and i hate it. and i know he loved this song:D but he choose to not sing this for a reason ... thirty four years desperately wanting "oh i wish michael would sing this! " 🤪 hear this and now i have.
 
Last edited:
zinniabooklover said:
Can't see an easy answer to this AI thing. Like any human invention it's going to have positive applications and also very negative ones, imo. From the i newspaper, 21 April 2023, I can find these stories:

1 - Liam Gallagher is happy with an AI project that has created a new 'Oasis' album. A band called Breezer has created it under the name AIsis. Liam is quoted as saying, 'It's mad as f**k, I sound mega ...'
2 - A German tabloid, Die Aktuelle, has printed an interview with Michael Schumacher using quotes generated by AI. You could argue that everyone would know those quotes are fake since Michael, afaik, is still in a deep coma. But it still sucks that they did that. Fgs, that's just gross.
3 - Big article on AI in the art world. Lots of arguments raging.
Otoh, you get this:
"Veracity Protocol say that the technology ... has the potential to be used more extensively in the future to authenticate genuine works and identify forgeries, establish provenance, provide valuations, and facilitate displays."
Otoh, there is also this:
Art historian / art dealer Bendor Grosvenor says, "I'm wary of anything which asks us simply to trust the machine ... I'm also wary of anything which begins, 'simply by taking a picture with your phone', Because it's more complicated than that. The authenticity of the Mona Lisa is as much about the materials that lie beneath the surface as the image which appears on top.'

Maybe none of that stuff is relevant to Michael. I can wrap my head around TII or demos being released but I can't get my head around AI. It just feels ethically wrong.


Ngl, I can't make this analogy work. People being influenced by someone's work is entirely different, imo.
Click to expand...
me too zin i still cannot make this analogy work ..... i have tried :(
 
Mister_Jay_Tee said:
I think most of us here can agree, regardless of our feelings on AI, or the ethics, that MJ would love this. At least be curious about it. Anger would be the absolute last thing on his mind.

Would he outright cosign it, like Grimes did? Maybe not. But after a while he might experiment with it. He had his own process for making music anyway.

I don't want an AI future of MJ music either, I'd rather hear the few complete tunes and then the demoes in a box set, then we can probably stop. Don't ever create AI music. And probably make duets out of the incomplete stuff, or tribute covers.
Click to expand...
When mf upload shit like this , when ..they use it this way which is happening all over yt .... I would believe that he wouldn't be happy at all.
You cant be happy hearing this right ? " madonna " it sounds possessed .
 
wendijane said:
When mf upload shit like this , when ..they use it this way which is happening all over yt .... I would believe that he wouldn't be happy at all.
You cant be happy hearing this right ? " madonna " it sounds possessed .
Click to expand...
Idk....I kinda like it😅😅
 
Those are making these AI MJ tracks are you using Paperspace to train your models? How many MJ acapellas have you trained on? What size are each file? Like 15-17 seconds? Lastly, how many epochs do you let it train for? Or how many hours/days/weeks?
 
wendijane said:
When mf upload shit like this , when ..they use it this way which is happening all over yt .... I would believe that he wouldn't be happy at all.
You cant be happy hearing this right ? " madonna " it sounds possessed .
Click to expand...
Obviously if it sounds like crap it's bad.
 
This is an interesting new field arising and it seems to be happening whether we like it or not. Since it's here I've looked into ai and since I'm an MJ fan I can't help but to listen to MJ Ai music. Some artists are spot on but every AI MJ song I heard drastically pales in comparison. MJ's understanding of rhythm, melody, tone, and emotion is something machines struggle to perfectly imitate. AI MJ typically sounds mechanical or sounds like an impersonator. If someone tries to use ai michael to appear as real michael leaks they are going to have a hard time getting people to believe it. MJ'S music is visceral by nature, producing that kind of music would require not just a soul but his soul. Only MJ can make MJ music.
 
Can someone make a ai version of tbis dr.freeze demo for mj pleaaaase:


I hear the magic ai version non stop
 
I just started training my own model, I’ma bout half done but here is a link to the Acapella of Music and Me from when he was younger, I redid it with his adult voice using the AI model I trained with Huggingface and Paperspace. Stil imperfect but absolutely getting there. I’m going to start again with my model and have a much larger data set as I only trained this model on about 7 or 8 acapellas. Would love to hear thoughts.

 
who wants to train their ai voice to do that (not michael though) does it take 90s gansta rap smooth and fluidly and mutiply voices so like can you train it to be all 5 ?
?
 
Dangerous1603 said:
how do you train the models? would you bravely explain, I've been trying for months
Click to expand...
I paid someone to train me and took me quite a while to get the hang of it, it’d be a bit difficult to explain it in a post. Sorry.
 
wendijane said:
You paid someone ? It's not a bedroom dj thing is it...
Click to expand...
Haha unfortunately not, although I’m sure before too long someone will turn it into software that’s easy to use. Right now it involves using Huggingface and setting up PiTorch on a virtual machine on Paperspace, setting up the training model, making your data set and then training on the virtual machine for hours, days or weeks. Basically you’re training your own neural network based on MJ’s voice, once you’ve got it trained, it’s pretty easy to use with some very very basic code. There is a guy on discord who taught me if you PM me I can give you his details.
 
wendijane said:
who wants to train their ai voice to do that (not michael though) does it take 90s gansta rap smooth and fluidly and mutiply voices so like can you train it to be all 5 ?
?
Click to expand...
To train it you’d need all five voices individually in acapella trained in different models then mix it yourself. I might be wrong but I can’t imagine it would do well with multiple voices in one track. When training my model, I had to each song into 15 second clips where it was JUST MJ’s voice, no background, no music or anything. Again I might be wrong, someone else might know how to do this.
 
Ie i‘m finding it hard to find clean Prince Acapella’s, to make my own Prince model so I could finally hear MJ and Prince on Bad. The fact there are so few clean acapella’s to train it on make this super difficult. Same with old artists like Elvis and Johnny Cash etc who recorded live with the band more often than not.
 
MKTCREATE said:
To train it you’d need all five voices individually in acapella trained in different models then mix it yourself. I might be wrong but I can’t imagine it would do well with multiple voices in one track. When training my model, I had to each song into 15 second clips where it was JUST MJ’s voice, no background, no music or anything. Again I might be wrong, someone else might know how to do this.
Click to expand...
thanks ..sounds technical ..very , The training , teaching it is the part which is fascinating ..
 
Last edited:
You must log in or register to reply here.
Back
Top