Offline
Cincinnati, Ohio

SO... I've been playing with some of these AI music tools and they are clunky but there is some amazing potential hidden in that mess.

HOWEVER

The use of these tools brings up a lot of questions. And there are a lot of different types of tools too which are not all made the same. Everything from open source basement-dwelling-hacker-code to the literal biggest technology companies in the world exists in this space.

I am totally open to the use of these tools under any color flag you want to fly them. In my mind, its a tool. It is just about the most complicated hammer I have ever seen, but it is still a purpose built machine with a usage. Lots of people of various dubious pedigrees make hammers. Is it ok to use a hammer trained on snuffing kittens? IDK man..

What do you think?

Offline
Bristol, UK

this is just gonna be my current thought on it, and its defo full of holes that I fall in all the time with my own stuff... but ultimately I feel if you don't understand how a tool, like AI, came up with the thing its done, it is not really 'yours'. So, if I got real smart and created an AI model that created music that I like, sure, that is mine. If I trained that AI on only Nine Inch Nails, then I think I might start having issues... If I use an AI model that I didn't make, and all I did was put a prompt in, I don't think I have any ownership of that, beyond a curiosity... But then.. If I started sampling that and creating my own work from that, then it becomes more an expression of myself again...

That's similar to how I feel about sampling, and how I feel about tools that write and suggest chord progressions, or even (the thing I use all the time) slicers and glitch plugins..


It has all these open questions, and I use ownership in both the 'I actually own the output of this so I'm free to share it' and the 'I can feel proud of the content I've made' way.

Ultimately, I do like the use of AI for prompting new creative acts. I ran a tabletop campaign recently, and in my depth of feeling uninspired, bashing "I need 20 missions for a pirate themed tabletop skirmish game" gave me no usable ideas, but I did get 8-10 things I could jump off of and use..

So yeah.. it is a biiig mess.. I don't think I'll use AI for music for a while, as I'm not a programmer, I don't know if I could ever make my own AI model, so I wouldn't feel like its 'mine', for whatever that is really worth. I can't wait to hear what people actually do with it creatively though!

Offline
Bristol, UK

oops double post.. classic

Last edited by alonemusic (August 21, 2023 6:54 am)

Online
St Louis

I been using random scales for years in electronic music. And others have been using Max patches for years that are generative. AI has already been around a while in music. And if you can't use AI in electronic music, where is it ok to use it? I'd be curious to hear what people make with it. Maybe put an asterisks by their names wink

Offline
Adelaide, Australia

Maybe it's a bit fatalistic, but there's not really anything we can do to stop it so I'm not sure its worth worrying about. There is an ethical concern when it comes to neural netwwork-based generation for sure, but its indistinguishable from pieces not using it, or using "traditional" random-number-generator-type generative techniques (which are typically quite popular in niche music communities like this - hello eurorack people!). The argument whether a piece of music is truly "yours" if you used another song's chord progression, or generated part of it, or accidentally used a melody that was in your brain but is only slightly different from one that already exists has raged for decades and will rage more.

The key thing, I think, is that there is no value to anybody in uploading a track that was rendererd wholesale from one of these new fangled "prompt a song" AI music generators, so I trust that people won't bother to do it (and those that do are lost to us already anyway).

Offline
Finland
Judgement Act wrote:

The key thing, I think, is that there is no value to anybody in uploading a track that was rendererd wholesale from one of these new fangled "prompt a song" AI music generators, so I trust that people won't bother to do it (and those that do are lost to us already anyway).

I kinda doubt that. I think all music related challenges will be flooded with prompt to audio "music" in same way that all art challenges are flooded with AI images. The problem is that, non-art people making AI art think their images are art as they think they look good. Same is with audio/music: People have generated really crappy sounding 22khz out of tune tracks and they really do not hear that it's not on normal level of music. They think it sounds great.

I'm very heavily into AI and have been running Stable Diffusion locally for 18 months. I have also been running AnimatedDiff for AI video. Also have trained many datasets with Dance Diffusion so i can generate endless amount of original drums / one shots. I recently did a kinda psych rock track where guitars, bass and vocals were AI generated and now one label is going to release it.

For me the potential lays in the non-muscality of AI. If i prompt list of groceries and AI tries to generate music from it, it can be really bad or something really interesting. Problem mostly is the too lo-fi sound quality. But that can also be advantage if looking for samples for lo-fi hiphop triphop.

Same goes for AI video. It can generate some really psychedelic stuff. Most boring what people do with AI video is those fake trailers where they generate image in Midjourney and then use img2video generator to make it move. I hope that was only 2023 thing.

Offline
Montreal

I think it is a very complicated topic. I wish AI would be mostly used to get rid of the tedious work that doesn't requiers "artistic" views. That we can use computers-robots to do the non creative tasks for us. I think it is all right to use computer as compagnions-tools to create as well but there are multiple ethical questions that comes with this.

I'll most likely try some of those new tools this year, probably will try to find some cool vocal generators, I already use auto-tune (with lots of me doing the work, but still, it is an AI tool I suppose). I use stretch audio to "fix" things up as well. I use LANDR "AI" Mastering too.

Never just feeded ideas to a computer and just used the output and I don't think I'll ever do that. I like to play instruments, to think of melodies and created more than "build" using already existing or generated lines.

I believe that with all the data available computer training is powerful and that it can generate very effective tracks by itself, but I like the idea of a human or group of human just sitting together and trying to build an idea-emotion, not selecting multiple elements of some of the music that did work good so far...

Offline
Coastal Australia
george bowles wrote:

I been using random scales for years in electronic music. And others have been using Max patches for years that are generative. AI has already been around a while in music. And if you can't use AI in electronic music, where is it ok to use it? I'd be curious to hear what people make with it. Maybe put an asterisks by their names wink


AI and generative music via the use of random/chaos/Max patches etc are very different thing imho.


If people are 'creating' tracks by only using prompts.... meh.

Offline
Melbourne

I've been using AI for the past year to help with lyrics, and it's been great. Obviously requires some editing to go beyond how generic the results often are, but it's lowering the barrier that I previously had in terms of finding words to sing on my songs.

In the past 6 months or so I've also been experimenting with various music generators as well. Suno is a good one that I've discovered recently - it not only generates music, but also intelligible lyrics and vocal melodies. Of course I often don't use the results directly - it's more like hearing a song on the radio and being inspired to use the chord progression or vocal melody style in a different context. However, it's also fun to get it to generate a stupid song I'd never think of writing and then re-record it using live instrumentation and vocals.. which I have done already this year and plan to continue doing.

Offline
Melbourne, Australia

I'm not too concerned by it, some music i hear definately contains the personality of the artist.... and sometimes it's just fun

I 100% plan to use A.I as a tool this year, in the sprit of experimentation.

The track you spent a few hours having FUN with, poured your self and ideas into vs over a prompt or a few cheap clicks means different things for both the listener and it's 'producer'. These are different outcomes/characters.


Relevant >>


Inspiration for good use of AI smile

Last edited by Aday (January 9, 2024 5:43 am)

Offline
Melbourne
Aday wrote:

The track you spent a few hours having FUN with, poured your self and ideas into vs over a prompt or a few cheap clicks means different things for both the listener and it's 'producer'.

Yeah I should say that in any case that I use AI, it’s just as a base/idea generator and I never use any of the generated audio.

Offline
Finland

This is pretty dope AI usage: i uploaded Mutable Instruments Shruthi1 manual to ChatGPT and asked it to come up some patch ideas:
1.    “Psychedelic Dream”
    •    Oscillator 1: zsaw, parameter: 64, range: -12, subosc: sq1
    •    Oscillator 2: fm, parameter: 32, range: 0, detune: -2
    •    Mixer: op: sum, osc bal: 64, sub osc.: 32, noise: 0
    •    Filter: cutoff: 40, resonance: 60, env1->vcf: 32, lfo2->vcf: 24
    •    LFO 1: shape: ramp, rate: 60, attack: 0, mode: free
    •    LFO 2: shape: sample&hold, rate: 52, attack: 0, mode: slave

No env etc settings, but it knows the weird oscilators of Shruthi.

Offline
Internet, Australia

it's like cheating at a board game while playing casually with mates. if you feel the need to do it, go ahead, but you'll always be the guy who cheated at board games, and if you can live with that, you probably need the excitement more than we do.

Offline
Arg

I sincerely doubt AI can come up with creative sound design, cool mixing techniques and anything that evokes profound emotion, so uhhhhhh.... I think it's a great set of tools to learn and use for help, but depending on it will get you nowhere