'Mixing revolution' ...

[Burillo]

We will have to wait and see how AI will affect music. I don't hold much hope for it. But I don't know if I will be around to see the result.

sure, but you not having hope for it is besides the point, because 20 years ago I don't think anyone could've hoped for the AI image generation we have now, or for ChatGPT. these seemed almost impossible because of how measly attempts at AI generation and AI art have been previously, but now you're like, oh, ChatGPT, that's yesterday's news and is no longer impressive.

don't get me wrong, I think there are some fundamental technical challenges that music industry as a whole has to solve before AI can meaningfully integrate itself into the workflow (basically, the entire workflow has to be reimagined to integrate AI into it - expose all the hidden and implied knobs, allow much more flexibility and more uniformity in terms of FX, etc.), so there's a really, really long road ahead before AI tools can even make a dent in someone's mixing workflow. however, for certain "simple" tasks (like, say, dialogue), AI tools like iZotope's mixing assistant can shorten the cycle already, and those tools are very rudimentary and are not very AI-ish to begin with. it's easy to integrate code-completion, because code-completion tools have been around for ages, and all you have to do is replace an algorithmic code-complete with AI-powered code-complete. currently, there is no way for a plugin to manage a mixing session even in an algorithmic way, so there are no tools we can use to substitute for AI. iZotope tries to get around it with inter-plugin communication, but this is vendor specific and very limited in scope.

in other words, the problem is more with integration than it is with the concept of using AI for music. once we get better integration between plugins and DAW's (something like ARA on steroids), then we can expect more and more AI-powered tools to appear, because they need infrastructure they can operate within - freestanding AI tool is not very useful, because the impact it can have is vastly limited.

[mevla]

Interesting discussion. Maybe a parallel can be made with MidJourney ? That would be classified as subjective, although I find that even a sunny landscape with flowers made by MidJourney has a gloomy feeling. As if in a corner would lurk some dark creatures, ready to jump in.

As for perhaps a practical approach, I recently tried the soundcloud mastering. Sure, there options, you can adjust a few things here and there, but the result, when not lacking in some ways, is kind of cold. SC (Dolby I think) mastering is certainly not the finest offer in town, though.

Then comes the question, why would I spend days learning how to operate an AI-driven approach instead of learning all the details of making the work using regular plugins and mixing, that can be arranged in hundred of ways thus offering a documented flexibility that can be put to work in many contexts ? Then based on the gathered documentation that I generated I can think about approaches - or approaches can be popping up - that are new, leading to other things.

The core is not the same. On way side there's the 'I', on the other there's less of an 'I'. So to speak. The person relegates some of his capabilities to working with an A.I., to adjusting A.I. parameters.

It's not a new debate. Back in the days when I did amateur 35mm photography I had a dark room to develop films. People were asking why don't you drop your rolls of film at the drugstore and be done with it ? Why do you give yourself all this trouble ? To this I replied that developing and printing on my own enables me to obtain exactly what I want to express and share with people. Deeper discussions can also include philosophy. Eg.: less of you means less of you sharing with other people, and subsequently, more of the A.I.-made presence shared with people.

[mevla]

[...] currently, there is no way for a plugin to manage a mixing session even in an algorithmic way, so there are no tools we can use to substitute for AI.

What are the tasks you are referring-to with 'managing a mixing session' ?

[Burillo]

[...] currently, there is no way for a plugin to manage a mixing session even in an algorithmic way, so there are no tools we can use to substitute for AI.

What are the tasks you are referring-to with 'managing a mixing session' ?

without thinking about it too much... more dynamic mixing templates (auto creation of send channels, sidechains, reverb buses, etc.), basic mixing (leveling, gating, tuning, etc.), that kind of thing - in other words, making sure that if a plugin wants to know what's happening on other channels and do something about it, it should be able to, and the DAW should expose all the relevant knobs to all plugins so that they can do what they will with your session.

for example, it could passively watch your work for a time, and then start making suggestions - like, "hey, this appears to be a vocal track, do you want to add things you usually use on your vocals, and set them up accordingly?". this sounds like it's something simple and something that is handled by templates and presets, but AI can take it to the next level and be more flexible, so that you don't have to go through the work of creating presets and generalizing your workflow - AI can do that for you. this is, of course, a rather silly and somewhat not-very-impressive example, but it's little things like these that AI workflow enhancers excel at.

[al808]

If AI could help with removing spill, string noise, fingerboard noises, deal with boominess and unwanted resonances in acoustic recordings, that's something I'd like to see. Cheers Al.

[studiowaveform]

If AI could help with removing spill, string noise, fingerboard noises, deal with boominess and unwanted resonances in acoustic recordings, that's something I'd like to see. Cheers Al.

Oh yeah that would be cool. I'd like latent space X/y pad with text input, so you could put acoustic, synthesized, clean, dirty in text prompts and move the pad and the AI would re-master the audio

[mibby]

Mixing Revolution Haiku *

Electric whispers,
AI orchestrates the blend,
Harmony transcends.



* courtesy ChatGPT

[mevla]

without thinking about it too much... more dynamic mixing templates (auto creation of send channels, sidechains, reverb buses, etc.), basic mixing (leveling, gating, tuning, etc.), that kind of thing - in other words, making sure that if a plugin wants to know what's happening on other channels and do something about it, it should be able to, and the DAW should expose all the relevant knobs to all plugins so that they can do what they will with your session.

[ ,,, ]

for example, it could passively watch your work for a time, and then start making suggestions - like, "hey, this appears to be a vocal track, do you want to add things you usually use on your vocals, and set them up accordingly?".

The "hey this appears to be a vocal track ..." can be highly, extremely annoying.

What you described can indeed be self-made as presets, or offered as flexible presets.

MAutoAnalyzer today does much more than that in terms of supporting mixing, technology wise.

When a possible A.I. is concerned in human creative domains I tend more to refer to what we can see now on Soundcloud/Dolby, as I tried recently : choosing an Aurora mix. Or a Misty Dew mix. Or others in the same vein. Make your selection, sit back and relax, the A.I. will take care of it. You can still afterwards run the resulting mix in say, MAutoDynamicEQ.

I still go back to philosophy regarding A.I. : the more an A.I. does in creative domains, the less an individual is sharing with others despite appearances . Hence, creations are exchanged in populations that contain a fair amount of A.I. actions. Hence, people are listening, consuming, art forms that are less reflective of what human beings are. This will have an influence of people.

In a similar context the European Commission has issued a document early 2021 that warns about using A.I. in social services. In this case the A.I. is specifically termed 'high-risk A.I.". Here the gov't has issued a $10 million support early 2022 to develop an A.I. that will be able to evaluate students in order to help over burdened teachers.

Social and art creation are not related, one might argue, although the receiver, the individual, is subject to the same.

[mevla]

If AI could help with removing spill, string noise, fingerboard noises, deal with boominess and unwanted resonances in acoustic recordings, that's something I'd like to see. Cheers Al.

This brings the question : how could people have done it until now and still some of them be highly famous and earning a good living ?

[Svama]

Too me, that's the wrong question. There are people who just love making music, playing instruments, but don't know how to mix there own song. They would really be happy. No one can and will learn everything.

[mevla]

Too me, that's the wrong question. There are people who just love making music, playing instruments, but don't know how to mix there own song. They would really be happy. No one can and will learn everything.

But isn't communication part of the whole thing ? I mean, there are ordinary people (charging no fees) that can mix and master. Or do it for a minimal fee. I never asked, but once there's someone in the U.K. who wanted to remix a piece I made for the studio he set up.

The other aspect is that, is A.I. a quick road to make money for simple folks who 'cannot mix and master' ? A bit like MidJourney combined with Red Bubble. Then, why not use 'tested and proven effective' riffs and rhythms made by A.I. ? There's already some music created by A.I. if I'm not mistaken.

It still boils down to sharing less of humans between themselves and more of A.I.

Why do I go back to that ? Years ago, during the times of popularity from film directors such as Pedro Aldomovar, it was observed in Spain that the country no longer had much of a mirror on their screens to reflect themselves. That being able to have a reflection of themselves - for any population - is a good thing to have. So there was a 'revolution' in the film industry that yielded many made in Spain movies, with some making it in foreign countries. Likewise, if at least the Western populations slowly relinquish the portion of themselves they are sharing with each other, then something else will take the place, A.I.

Again, philosophy.

[E_Anderson]

Hence, people are listening, consuming, art forms that are less reflective of what human beings are. This will have an influence of people.

This is exactly how I feel about it.

"Trying to add music to your next video? Then you need to check out Artlist!"
This is an ad I get pretty frequently on Youtube, and I can't stand it. The specific thing I don't like about it, which is not unique to this ad, is the use of smiling - the voiceover artist is being told to smile widely and convey a lot of happiness and excitement about Artlist when mentioning it to the target receipient. It's just a lie, emotional fluff used to make something out of nothing and try to get people to buy their product.

It's not to say they don't have to do it; if you're trying to make it big with a product like that, most every market is saturated and you have to pull out all the stops to get your business noticed. But that kind of "well these are the attributes of communication that people respond to, so let's apply them everywhere we can to maximize engagement" appears everywhere, in ads and in popular music and it's very... Algorithmic, not human. It doesn't feel like a person. And that person-value I agree is being eroded year by year.

I feel like none of this made the kind of sense I would have liked, and this is also taking place in the Mixing Revolution thread, where I originally came just to see if there was any news + maybe speculate a bit on what kind of workflow it would offer (something like MIR where inputs are arranged spatially on a 2D plane, and then effects can be applied to each voice in a more visual way than mixer-track-navigation kind of way?), but here we are philosophisiphizing about societal decline. AI and its implications is a topic... routinely on the mind these days