AI in mixing/Mastering/Composing

I think its depressing. And scary. Someone without any skills or experience can soon (now?) make a hit song and mix it, all with AI. Isotope 11 is great I have tried them, and the work is so fast. There are software that produces chord progressions with AI and probably does so very good. And you can sing terribly, then clone someones voice and it sounds really OK. And things will get much worse (better?). This is catastrophic for music, for mankind. Or what do you think?

I think this topic has already been done. More than once.

Anyway …

3 Likes

Yes, does that mean that the discussion is over? Its coming new software all the time now, thats why I picked this up again. This was in my email the other day.

I think it will kill the joy in making music, I agree, it gives me the creeps to.

I just downloaded Ozone 11. Haven’t used it yet (already running Ozone 9). But to me this makes a lot of sense. I want to focus on composing and performing. I had thought a while ago that it should be possible to simulate sending your mix to say Chris Lord-Alge with an AI mastering software, and it’s getting there. So sort of a shortcut to hitting the mark in making your mix sound like whatever is the current fashion in terms of loudness, punchiness, frequency response, width, and maybe even mimicking analog characteristics. I personally don’t feel like struggling with these secondary factors (secondary to composition and performance).

I think the discussion of AI composition and performance is a completely different topic. AI compositions can only be derivative, to my understanding of how AI works today. Originality is not remotely possible. At least not yet, although I don’t see any developments working in that direction. It’s more like, make a piece in the style of Mozart or the Beatles.

And then there is the question of acceptance of music that sounds original. That’s on us, the humans who would listen to it. For example, the people who heard something that moved them in Rite of Spring, for example. It’s a form of communication, isn’t it, from Stravinsky to his audience, even if we are not at all sure how that communication is working. It’s a kind of pulling or stretching into new musical territories, and not all are pulled there. But the pulling and stretching is being done by a human (Stravinsky), and the stretched and pulled are humans.

I wonder if that communication that I just described will ever be possible between a learning piece of software and a human. The learning piece of software, at this point, can only imitate. Nothing I am seeing convinces me that this will ever evolve.

1 Like

I guess to be fair this topic is a more broad conversation (?) than the other one which is more a discussion about specific AI that people actually want… ?

I think true artists will stay away from that stuff.

It’ll put a fair amount of music mixers out of work in the process though. And once it comes to the stuff you feel like struggling with they too will be replaced sooner or later.

That’s not how I understand it. But even if that’s the case, so are most creators - derivative. Hearing something that is actually conceptually new is really unusual.

Think about it this way; in the previous century we saw Jazz created and its early forms split into be-bop and its offshoots as well as more avantgarde freeform jazz and fusion. All quite distinct. We got Rap and Hip Hop. We got Rock music which split into say Hard Rock and Punk Rock. And Disco and Electronic music like House Music.

But in the past decade what’s really been new? To me it seems like every time I see a new label and I start reading up on it and listening to artists in the “genre” the distinction between them and others is tiny. So the idea that something is “new” has changed because the difference is much smaller. It’s more like sub-genres.

So anyway, what can we create today that hasn’t been done already? Someone might come up with something but I’d say it just doesn’t seem it will be likely to be as profound as the new genres of the last century. And so with that in mind will a ‘derivative’ AI really be any worse than a derivative human? I mean, I was right in the middle of the 80’s, and how many 80’s “revivals” have we had now? I love the sound but every time someone new does that again today I’m yawning. Same with funk music. They do all the right things the right way but without the actual funk.

And going back to the AI and how it functions: I would think that an AI should be able to also analyze prior developments in music, ones we’ve liked, and understand what differences we find acceptable and simply apply those procedures to create something new. So while we might think that it’s derivative because we say “create something like Mozart” what could be derivative is the concept behind creating something new which isn’t unlike how we do it (if it learns from us anyway).

Lastly - it’s going to suck for all music producers and composers out there. Those that do it for a living. AI is going to kill the middle of the market and downward.

But all musicians imitate? Whe use 3 or 4 chords for many, many songs, music is not that complicated? I have used Ozon11 also for the same reason as you, Clarity is fantastic. And that is different from composing. I would prefer to not use AI for mixing but I am also a singer/musician and want to save time on mixing/mastering. An AI can spit out hundreds of chord progressions in a few minutes, then a human can choose the best one so even chance is in favor of AI. and chance has created many inventions, and also a lot of good art.

I agree, 100%. AI will be the downfall of music and musicians. We already see most the new album covers on spotify are made with Midjourney, an text to image AI.
I got one new AI plugin today in my mail, it was also for making chord progressions.

From what I have learned about machine learning, it finds patterns in the training material which it uses to create new material. It doesn’t have the capability to develop new patterns. Of course that may change, but will the new patterns make any sense to humans? I guess we’ll see.

I don’t think AI can kill something that is already dead; thanks Spotify! The advent of Spotify might in itself be enough to explain lack of new genres. On the other hand, you can’t see new genres sometimes for decades. For example I didn’t know hip hop existed until the 90’s, yet it was supposedly born in the early 70’s.

I do hope that all the creatives won’t give up. Actually, I’m sure they won’t. And I’m pretty sure they won’t be superseded by AI any time soon. At least not on the cutting edge. As I said earlier, it’s a form of communication. From someone who experiences feelings to someone who can experience feelings. AI can’t do either end of that transaction.

It can create new patterns if you throw in some randomization.

If you’re talking about music as an artform, like bands and singer-songwriters that make music for their/its own sake then I don’t think Spotify is the whole story.

Music started losing its value when people started copying and sharing digital, and another huge blow was piracy. So Spotify emerged as a convenient low-cost alternative to having to download stuff. But some people still do the same for movies and TV shows - just download ripped stuff off of torrent sites.

People are selfish and cheap and don’t care about compensating other people which then reduces net income into the industry.

On top of that we have actual democratization of the means of production, meaning “everyone” can get a computer and record some music.

However;

For a long time artists also survived by having one foot in the ‘commercial’ world. So if you were a pretty avantgarde artist by night you might compose for actual advertising during the day. This field, which was comprised of excellent musicians and composers have been reduced to something many more people can do today thanks to technology, and AI will largely completely kill that. So ten years from now nobody is going to pay much for humans to create that, thus killing off a revenue stream for actual artists and also for ‘just professionals’.

It’s about getting help with everything from “Suggest a melody” to “give me the chords” to “hook me up with some instrumentation” to “mix it please” to “master it”. I mean, if we can ‘get help’ from AI literally at each step of the process - and it makes sense for the composer to get help with mixing and the mixer to get help with composition - then the AI that helps with all steps might as well just do it all.

It “dumbs” down things. Great for people wanting to be hobbyists and just play around, but generally bad. It devalues the massive effort (and/or talent) of the great artists of the past because the new normal is that anyone can do it… you just need the tools (AI)…


Having said that I think real live music will always have a place. Just maybe niche, unless we change society.

Yes, its up to the consumers, but we have the ABBA avatars and others are are following so even live music is in harms way. If you look at AI images I think you can see that AI can generate new stuff, it combines old stuff in a new way, and thats largely what musicians do to.

AI is rubbish, in mixing, mastering, and writing songs.