AI in music

[Lozz196]

For me it’s the imperfections or going where wouldn’t be expected that make music interesting, such as Cliff Williams playing C# in Back in Black when the guitars play A. Would AI choose to do this?

[SteveXFR]

I give up on Facebook and Instagram. AI has just made them even more irritating than they were before 

[TimR]

21 minutes ago, Lozz196 said:

For me it’s the imperfections or going where wouldn’t be expected that make music interesting, such as Cliff Williams playing C# in Back in Black when the guitars play A. Would AI choose to do this?

 

YouTube creators create A and B versions of titles and thumbnails to see which gets most attention. 

 

Maybe AI will do similar, or maybe it will the be humans' job to listen to tracks for quality control. 

 

Ultimately AI has no sensors or hormones to measure whether what they produce is working. 

 

Comedians and live musicians get instant feedback from audiences. Whether or not they listen to the feedback or not depends - there are plenty of artists who have spent years looking for audiences who 'get' them. 🤣

[chris_b]

I'd like to see anyone use AI to copy how I play. Good luck to them.

[SteveXFR]

1 minute ago, chris_b said:

I'd like to see anyone use AI to copy how I play. Good luck to them.

 

Maybe thats the way to sabotage AI, expose it to the out of tune noodling of the average BC member. I could probably break it myself. 

[Nail Soup]

For doing creative tasks - no thanks. 

But can be a great tool for things like separating music into stems or e.g cleaning snare rattle from an acoustic guitar passage or whatever.

 

[BigRedX]

14 minutes ago, SteveXFR said:

Maybe thats the way to sabotage AI, expose it to the out of tune noodling of the average BC member. I could probably break it myself. 

 

Poisoning the well.

 

IIRC someone has created a method of encoding music so that AI thinks it is a different genre to its actual one, or "hears" the music it is being trained on as just noise.

[BigRedX]

Right now for creative purposes. AI is at the same stage I was in my teens when my songwriting "influences" were very obvious, and the greatest current threat to it is that it will get sued for plagiarism.

 

As others have said there are lots of uses of AI in music that take the donkey work out of some technical aspects such as noise removal. and frequency balancing.

 

I use AI fairly frequently in my day job in graphic design. One use is when creatinge photo-realistic mock-ups of some of the food products and packaging that my clients are considering producing. I could spend time in Photoshop compositing inclusions onto a slab of chocolate to show through a window in the packaging, or I could ask an image generator to create one for me. One will take an hour or so the other with the right prompts less than 5 minutes. 

 

It has also been used to generate one of the "covers" for my band's releases. At the time I was completely snowed under with paying work, so we fed a series of prompts into an AI image generator and refined the output until we had something everyone was happy with. I could have done a better job manually, but it would have taken a lot longer and at the expense of paying work.

[Al Krow]

Scarily easy as I've just figured out! 

 

Another nail in the coffin for aspiring originals artists hoping to "make it", just as streaming has already proved to be?

 

I suspect that where the industry is heading could well be similar to where classical music has ended up with performers, in our case covers and tribute bands, being custodians of a vast library of music that is already much loved; with occasional new material to sit alongside the great songwriters and composers of the past: read ABBA to Zeppelin for Bach to Verdi.

 

The Times They Are A-Changin'...

 

Edited 2 hours ago by Al Krow

[SumOne]

28 minutes ago, Al Krow said:

Scarily easy as I've just figured out! 

 

 

Another nail in the coffin for aspiring originals artists hoping to "make it", just as streaming has already proved to be?

 

I suspect that where the industry is heading could well be similar to where classical music has ended up with performers, in our case covers and tribute bands, being custodians of a vast library of music that is already much loved; with occasional new material to sit alongside the great songwriters and composers of the past: read ABBA to Zeppelin for Bach to Stravinsky.

 

The Times They Are A-Changin'...

 

Yeah, I figure playing an instrument live is somewhat future-proofed. People have been able to listen to recorded music for over 100 years but still want to go out and hear people perform it. 

 

Making new music, I dunno - AI will get very good at it. There will probably be a lot of hybrid stuff where music is made more cheaply/easily with AI but there is still a real 'face' to the music e.g. ask AI to make a backing track after having been given prompts about the style etc. but then a real person sings over it.

 

 

 

Edited yesterday at 16:41 by SumOne

[Al Krow]

2 minutes ago, SumOne said:

 

Yeah, I figure playing an instrument live is somewhat future-proofed. People have been able to listen to recorded music for over 100 years but still want to go out and hear people perform it. 

 

Making new music, I dunno - AI will get very good at it. There will probably be a lot of hybrid stuff where music is made more cheaply/easily with AI but there is still a real 'face' to the music e.g. ask AI to make a backing track after having been given prompts about they style etc. but then a real person sings over it.

 

Even more scary is that a close mate, whom I formed a covers band with back in 2013 and has written 5 albums, just commented: "TBH that song sounds as good if not better than many songs coming out of Nashville".  Whether that's fair or not, the fact that was his reaction just shows how good AI already is at composition.

 

Very glad that, like you, I'm focussed on live performance. 

[Stu-khag]

My colleague made this programme (I did the logistics for it) .

Worth a listen.

 

 

As a composer in San Francisco, Tarik O'Regan is at the epicentre of the collision of classical music and AI. In San Francisco now, there's the same buzz as in the early 00s' tech boom. But 'old tech' is moving out and Silicon Valley has a new nickname: 'Cerebral Valley'. Around Tarik's studio, buildings are constantly bought up for AI start-ups and ‘Hacker Houses’ - where budding engineers live, eat, code together - proliferate. And one of AI's fastest areas of development? Music creation. But classical music is part of San Francisco’s life blood. Since the 19th century, it's had a reputation for concert halls and opera houses. British composer Tarik O’Regan takes a personal journey around the city he now calls home, riding a driverless taxi to talk to figures from both the tech and classical music worlds. Will AI ultimately turn his professional career into a hobby, or will it prove a useful tool to human composers? He talks to Tamara Rojo from San Francisco Ballet, composer Mason Bates, and composer and AI entrepreneur Ed Newton-Rex, and uses AI to compose a new work to be performed by the BBC National Orchestra of Wales.

 

https://www.bbc.co.uk/sounds/play/m002k3th

[Marvin]

Within music (or other arts) I have no interest at all in AI content. It's pointless oxymoron, that try my upmost to avoid.

 

I find it ultimately boring. There's no 'story' so to speak about the creative journey, bar a few prompt words.

 

I've got to the point where I'm bored rigid of nearly every movie being a CGI fest, so AI has no chance. 🙂

 

AI is able to step in because unfortunately creativity is being discarded in so many areas.

Music, a lot of uninspiring dull generic output. Movies, constant remakes and sequel franchises. 

 

I'm off to my cave, cheerio.

[TimR]

21 minutes ago, Marvin said:

I've got to the point where I'm bored rigid of nearly every movie being a CGI fest, so AI has no chance. 🙂

 

Quite. The CGI should enhance the storyline, not be the whole film. 

 

AI in music should be there to do the same. Improve, not replace. 

[Al Krow]

Been a very interesting 24 hour journey for me. This thread (and thanks to the OP for starting it) prompted me to look into what the AI song-writing process involved, which I'd not done before. And as I commented earlier - it's scarily easy!

 

I shared the song earlier on this thread, and also with a few close muso friends, who were very complimentary about the quality of the song.

 

One mate, who's also a bass player and band leader of a covers band, rightly challenged me by asking: where's the "pride, value, wonderment" in using an AI song writing platform? He's 100% right that it would feel pretty shallow in itself, but as part of the full YT video which it inspired, well quite a surprising amount! It's probably the most creative thing I've done in a long while!

 

 

Will be fun to see if we can play this well, live, at some point down the line!

 

Edited 2 hours ago by Al Krow

[EBS_freak]

2 minutes ago, Al Krow said:

Been a very interesting 24 hour journey for me. This thread (and thanks to the OP for starting it) promoted me to look into what the AI song-writing process involved, which I'd not done before. And as I commented earlier - it's scarily easy!

 

I shared the song on this thread and with a few close muso friends, who were very complimentary about the quality of the song.

 

One mate, who's also a bass player and band leader of a covers band, rightly challenged me by asking: where's the "pride, value, wonderment" in using an AI song writing platform? He's 100% right that it would feel pretty shallow in itself, but as part of the full YT video which it inspired, well quite a surprising amount! It's probably the most creative thing I've done in a long while!

 

 

Will be fun to see if we can play this well, live, at some point down the line!

 

Did you prompt to be country - Im amazed how much AI music output is deeply routed in country and country cliches. To me, AI seems to excel at country and soul - probably bringing out the bias of the music its had in it's training set.

[Al Krow]

1 minute ago, EBS_freak said:

Did you prompt to be country - Im amazed how much AI music output is deeply routed in country and country cliches. To me, AI seems to excel at country and soul - probably bringing out the bias of the music its had in it's training set.

 

Yes, exactly correct.

 

[EBS_freak]

So you didn't prompt it to be country?

[Al Krow]

12 minutes ago, EBS_freak said:

So you didn't prompt it to be country?

 

Nope "you being exactly correct" = I did prompt it to be country.

 

(ChatGPT would have understood my response 😅)

[EBS_freak]

Ah - what I am saying, is if you dont prompt a genre, theres a considerable chance it will come out as country... especially if the lyrics are US nostalgic.

[Al Krow]

29 minutes ago, EBS_freak said:

Ah - what I am saying, is if you dont prompt a genre, theres a considerable chance it will come out as country... especially if the lyrics are US nostalgic.

 

Interesting. Is that partly because Country is a relatively uncomplicated genre in terms of song structure and chord progression, and therefore a relatively easy one to get under AI's belt?

[dave_bass5]

I find learning songs is a lot easier now we have decent stem splitting, I use it all the time. 

 

I also can't sing to save my life, but have musical ideas that would just be pointless without a singer on them. Again, Ai is a big help with creating vocals to go on my instrumentals. 

 

 

[SteveXFR]

Something I do enjoy that AI couldn't replicate is when musicians try something weird. Like a band I saw at the weekend with a funk metal song about being a fish farmer or everything Igorrr has ever done.

[EBS_freak]

1 hour ago, Al Krow said:

 

Interesting. Is that partly because Country is a relatively uncomplicated genre in terms of song structure and chord progression, and therefore a relatively easy one to get under AI's belt?

Nope. It's cos it's what the training data comprises of.

 

To emphasise the point, if 80% of your training data was to be country, it should come as no surprise what the output sounds like.