Here's what we feared: the entirely AI synth - guk.ai

[BONES]

those who are pro machine creative assistance i'm sure could provide numerous scenarios for justifying the use (many are bound to be rationalizations, others may in fact hold water in a given context), but the inherent hack-ness of the basic idea is probably hard to escape, as it should be

Why would anyone need to rationalise a perfectly rational decision? There are plenty who'd tell you that using a DAW is cheating but I'm sure you don't feel the need to rationalise your use of a DAW, do you? AI is a tool like any other, you use it when it's the right one for the job. To be fair, though, this thing seems like a bit of a waste of time.

When I think of using AI, it kind of reminds me of those large people in the last section of Wall-e. They don’t do anything themselves, drinking big gulps, and just sit around all the time.

That's because you've never actually tried to use AI to get anything done. It's hard f**king work.

[BBFG#]

Using an AI is "hard f-kn work"?
That's the best reason to avoid it I've read.

[starflakeprj]

If you've chosen not to embrace AI as part of your everyday life, you'll soon be left behind, I've read.

[Examigan]

When I think of using AI, it kind of reminds me of those large people in the last section of Wall-e. They don’t do anything themselves, drinking big gulps, and just sit around all the time.

Then you're doing it wrong. If you are in information work of any sort and you haven't figured out by now how to use LLMs as a force multiplier then perhaps the lack of intelligence in AI isn't the correct diagnosis.

I just meant using AI to create presets for synths seems lazy to me.

[forgetitbestie]

If you've chosen not to embrace AI as part of your everyday life, you'll soon be left behind, I've read.

Yeah and the person who wrote that is a rube or has a personal stake in AI success

[starflakeprj]

Does it matter though? It’s still inevitable, still true.

[vurt]

100100110100101001

[billinder33]

That's because you've never actually tried to use AI to get anything done. It's hard f**king work.

How are you using it?

[coffee or coffee]

those who are pro machine creative assistance i'm sure could provide numerous scenarios for justifying the use (many are bound to be rationalizations, others may in fact hold water in a given context), but the inherent hack-ness of the basic idea is probably hard to escape, as it should be

Why would anyone need to rationalise a perfectly rational decision? There are plenty who'd tell you that using a DAW is cheating but I'm sure you don't feel the need to rationalise your use of a DAW, do you? AI is a tool like any other, you use it when it's the right one for the job. To be fair, though, this thing seems like a bit of a waste of time.

When I think of using AI, it kind of reminds me of those large people in the last section of Wall-e. They don’t do anything themselves, drinking big gulps, and just sit around all the time.

That's because you've never actually tried to use AI to get anything done. It's hard f**king work.

i use the word rationalize because, in my estimation, the negatives of ai use far outstrip any positive aspects. from the devastating environmental impact to how its trained by plagiarizing the arts to how its being applied and utilized as a business strategy and just how error prone it is, that shit, for me, doesn't vanish because it has some minor advantages in certain circumstances.
i find the use of a such a malevolent little greed fulcrum to be irrational, so in my scenario one that would need rationalization.

and to be fair, i was extrapolating the use of ai from this shitty plug to a wider more involved use of creating art. i dont see it as a tool to make art, its a tool thats learned by stealing to create a homogenized version in response to key words.

while i get the point you make about the daw angle, possibly a more appropriate example would be to compare it to sampling maybe? even that is only surface level, because there are miles, thousands of miles of difference between those things and ai

[ghettosynth]

When I think of using AI, it kind of reminds me of those large people in the last section of Wall-e. They don’t do anything themselves, drinking big gulps, and just sit around all the time.

Then you're doing it wrong. If you are in information work of any sort and you haven't figured out by now how to use LLMs as a force multiplier then perhaps the lack of intelligence in AI isn't the correct diagnosis.

I just meant using AI to create presets for synths seems lazy to me.

For you, perhaps it is. For someone who's strength lies elsewhere and they just see that as a routine part of what they have to get done, not so much. To be clear, the product isn't appealing to me, but that's not because it's AI driven, it's that it's not the kind of AI product that fits into my worldview, let alone my music workflow.

That said, this differs from the premise in Wall-E, where automation consumed life until humans lost the will to create meaning and waited for purpose to return once everything was fixed. In our world, AI isn’t the real analogue to that automation; social media is. It doesn’t take choice away; it breaks focus. It trades steady effort for constant hits of stimulation, replacing creation with reaction. Wall-E imagined paralysis. We live in distraction. Meaning isn’t gone, just hidden under noise and reward. Even so, it can still serve a purpose if kept in check.

Generative models are incredibly useful. Don't confuse that fact with the false notion that everything that people want to build with them to extract your attention and money is also useful.

BTW: Wall-e is a fantastic movie and is on my rotation for when the dementia sets in and I'm just watching old movies over and over again on the couch.

[BONES]

How are you using it?

To write songs for our next album, or maybe as a side-project (if we decide to keep the AI vocals). As I've said several times already, if you just want to flood Spotify with a millions shitty songs, AI can do that at a rate of knots but if you want to craft something specific, you have to work at it. You have to think long and hard about your prompts, trying a number of approaches until you start to get what you're after. Once you have something with potential, you have to iterate and revise to finesse it into shape. We probably end up with one usable thing for every 500-600 attempts.

With both of us putting in the same sort of hours we've put into previous albums, I expect it will take us two or three months to have an album's worth of usable material. That's probably faster than we'd normally get an album together but with the AI generated stuff, we have a lot more work to do once we've got the songs. We'll probably try doing it a few different ways but my first attempt will be to rebuild the songs in Studio One, using VSTi and my vocals. We'll also try to do stem separation, then cut up the different sections to rebuild the arrangement without the artifacts you get from straight stem separation. If neither of those things work, we'll just play around with the AI generated material in Ozone, to rebalance the mix (the AI vocals are always too loud), maybe add a few extra bits and release it as a side-project.

At the moment we've got 3 or 4 things we can use, things we want to turn into NOVAkILL songs, plus a few others that are great but not for NOVAkILL. It's pretty much all quite listenable, though, so if we wanted to we could probably do an album a week, put it up on Bandcamp and make money from it. But that's not what we're interested in.

We want to make the best songs we can. 12 months ago that meant mostly doing it ourselves in Studio One, with a bit of lyrical assistance from AI. Today, however, we reckon AI will give us the best songs overall, as long as we're willing to put in the effort. Even if it doesn't work out as we'd like, we'll still have an album of AI generated songs that we'd be happy to listen to over and over. That's the thing, to me it feels exactly like stumbling upon a new band that's making music that's perfectly aligned with my musical taste, which is something that happens very, very rarely these days. (Rarely as in twice in the last 25 years.) Interestingly, I also think it will have more commercial potential than the stuff we do ourselves, so it could get something we're proud of out to a much wider audience, which can only be a good thing for the music we love.

i use the word rationalize because, in my estimation, the negatives of ai use far outstrip any positive aspects. from the devastating environmental impact to how its trained by plagiarizing the arts to how its being applied and utilized as a business strategy and just how error prone it is, that shit, for me, doesn't vanish because it has some minor advantages in certain circumstances.

None of that is relevant to this discussion and most of it is bullshit anyway. e.g. AI learns pretty much exactly the same way we do so if I can listen to music for free and use that experience to make my own music, why shouldn't AI be allowed to do the same thing? And how would an AI be "wrong" when making music? Honestly, take a look at what you're saying and see how it could be applied with equal validity to computers and the internet in general. It's just an evolutionary step along a continuum we've all happily been exploiting for decades. Why draw some arbitrary line now? It makes no sense.

and to be fair, i was extrapolating the use of ai from this shitty plug to a wider more involved use of creating art. i dont see it as a tool to make art, its a tool thats learned by stealing to create a homogenized version in response to key words.

In the words of Picasso, "good artists copy, great artists steal". So what you're really saying is that AI does what the great artists all do, so how does that not mean that it is the perfect tool for the creative arts?

while i get the point you make about the daw angle, possibly a more appropriate example would be to compare it to sampling maybe? even that is only surface level, because there are miles, thousands of miles of difference between those things and ai

Sorry but I'm not seeing it. I see it more like a sample-based instrument such as Ujam's Striiiings, for example. That takes samples of real performances and allows a dumb c**t like me to have full, lush orchestral arrangements in my music without having the slightest clue how to actually do that. Who is going to ignore an advantage like that?

[billinder33]

How are you using it?

To write songs for our next album....

Are you using AI to do arranging? Or just parts, acapellas, samples, etc? Where does the generative AI end and your other instruments begin?

Have you tried Suno Studio? I've been thinking about it as a potential companion to Splice (I'm guessing Splice will be or already is majority AI content anyhow). Also for prototyping arrangement ideas, though not necessarily using the audio in that context.

[BONES]

Tunee, the AI we're using, spits out complete songs with vocals, which is why stem separation is necessary. Because of the iterative nature of the process, you end up with lot of versions of the same thing, so cutting the good bits from one and pasting them into another is one of the things we do to get closer to what we're after. The problem with that, though, is that it never does it at the same tempo twice, so there's a fair bit of faffing around to get it all working and sometimes the easiest thing is just to work out how to play the part yourself and play it into the sequencer.

To be honest, the good ones are fine the way they are but we can make them better with a few tweaks. It's terrible with lyrics, you can tell what you want but it will do it's own thing if it feels like it, which is not a problem for us as we're not going to use its vocals anyway, unless we decide to do it as a side-project.

[CinningBao]

what a fkin palava. you think you're thinking useful stuff and learning things.
what are you actually learning?

[billinder33]

Tunee, the AI we're using, spits out complete songs with vocals, which is why stem separation is necessary. Because of the iterative nature of the process, you end up with lot of versions of the same thing, so cutting the good bits from one and pasting them into another is one of the things we do to get closer to what we're after. The problem with that, though, is that it never does it at the same tempo twice, so there's a fair bit of faffing around to get it all working and sometimes the easiest thing is just to work out how to play the part yourself and play it into the sequencer.

To be honest, the good ones are fine the way they are but we can make them better with a few tweaks. It's terrible with lyrics, you can tell what you want but it will do it's own thing if it feels like it, which is not a problem for us as we're not going to use its vocals anyway, unless we decide to do it as a side-project.

I sent some time with Suno Studio last night to get a feel for it. Created a song, separated some trap vocals into a stem and pulled them into Bitwig. Suno will time align to a tempo aligned grid before export so that's a nice time saver. Didn't really use it for the arrangement, because I didn't drive it correctly to get something interesting. There is separate input for lyrics and music and I used the lyrics section to describe the entire song structure, which a very was basic structure on my part. There is also a section to describe the song in which I only added a few style tags. It output 4 songs which sounded almost the same, but maybe that's because I over described the song in the lyrics input.

It has built in stem separation so that was also a time saver, but I can see how this process could be quite time consuming depending upon how much of a finished product you want. My main interest would be vocals since IMO this is the biggest failing of tools Splice (decent acapellas that fit an existing track take forever to find, if you even find something at all when could already be copyrighted by someone else with you knowing).

Also not a cheap tool at $24 a month. Factor in and Splice and the Bitwig update plan and you're at close to $750 annually. Glad I'm already loaded with plugins and don't need any more.

Still haven't tried the synth in the OP. Lol.