SUNO is killer!

[VitaminD]

moving this post..

[zerocrossing]

I finally tried one of these, but not Suno. It billed itself as a better sounding alternative to Suno. My prompt, "An aggressive prog rock anthem about sleeping cats in the style of Adrian Belew." The results were interesting. At first listen, pretty impressive, but then I noticed that they just fell flat after a bit. Listening to each iteration was similar. Totally lacking nuance and never building to anything. Maybe Suno is better? Not sure. Also, the sound quality isn't good. I don't hear anyone complaining about that, but it's not good. I am interested in Ace Studio, though, as a way to get orchestrated bits done in a quicker way than I'm currently doing using Kontakt. I know it's on the ropes, but it would be nice if Kontakt actually did that. Let me play a guitar line into it and have it spit out a horn section with all the articulations done. I want help making music, like George Martin helped the Beatles. I don't want something to make music for me... we already have that. It's called, "people."

[zerocrossing]

The artifacts are so horrible it takes me one to two seconds to hear if something has been generated with AI. The best compositions made it into the meh category - nothing interesting, nothing I haven't heard a thousand times before, just with even less tension/release. The instruments are useless, the timbres are all over the place, constantly morphing. Performances are robotic, it's not possible to replicate human push-pull at all. The mixes are ridiculously bad with unstable transients, weird lowpass filter effects on cymbals and collapsed stereo fields (Holes between L, M and R). There are tons of serious issues. Suno is not a sheriff, not a deputy - it's that smelly weirdo with developmental delay who sleeps in a barn, unable to speak, only repeating words you say to him. He might be harmless but it's still not a good idea to make him the sheriff, he's likely to blow his head off by accident five minutes after you gave him a revolver.

This is also my take on it. It's not ready for the big time, not by a long shot. Not if you actually care about how something sounds. It seems to be OK with a single thing at a time, but when it tries to mix a bunch of things together, it's a weird sounding blend of mush.

[zerocrossing]

As expected, no demonstration. Because you're shitting your pants.

He's trolling you, sir. For comedy.

[zerocrossing]

None of the examples would qualify as good.

But they’re all considered good enough, and apparently that’s all that’s needed.

Good is highly subjective anyways. AI-generated music will never please anyone who is fundamentally against AI.

If you get a user error rate of 100 %, maybe the tool is the problem, not the user...

Maybe it’s not the one-click solution everyone is dreaming of. AI still requires you to do your homework, which is why I’m not worried in the slightest. You will still need a musician to make music, even with AI. It just lowers the barriers a bit, much like Music Maker did before, or the introduction of digital DAWs, or the invention of VST instruments.

Sure, lowest common denominator always works in the world of Wonderbread tastes, but it's one thing to say you don't like something, and another to say it's not good. For instance, I hate The Grateful Dead, but it's very good music. I could go on all day about how funny it is to me that the Italo Disco crowd spends hundreds of thousands of dollars on vintage gear to make what I find to be the dumbest form of dance music to ever exist on this planet.

What I'm talking about is how well can it mimic something complex, like an orchestra playing the Game Of Thrones theme, or something similar. Can it sound like it was recorded in a studio designed for classical music recordings? No. I guess that might be OK for you, but it is not for me.

[ksandvik]

'Wonderbread' is a good name for AI slop music, will start using it!

[ghettosynth]

Some like writing music like filling in spreadsheets...

Konstipatio VonVondervale Rumpelrauch does, it's his whole thing.

[ghettosynth]

What I'm talking about is how well can it mimic something complex, like an orchestra playing the Game Of Thrones theme, or something similar. Can it sound like it was recorded in a studio designed for classical music recordings? No. I guess that might be OK for you, but it is not for me.

I've decided that I should communicate more with music. So, this track conveys exactly the same thoughts as what I posted before.

[zerocrossing]

I know I keep bringing this up, but you all need to listen to the Shell Game podcast. Season 2. It’s all about a guy who makes a company out of all AI agents and himself. Their product ends up being an AI driven application that surfs the web for you, leaving you extra time to not have to deal with that. “Sloth Surf.” The agents even named it. It’s hilarious and indicative of where the technology actually is, and what’s hype.

[ghettosynth]

I know I keep bringing this up, but you all need to listen to the Shell Game podcast. Season 2. It’s all about a guy who makes a company out of all AI agents and himself. Their product ends up being an AI driven application that surfs the web for you, leaving you extra time to not have to deal with that. “Sloth Surf.” The agents even named it. It’s hilarious and indicative of where the technology actually is, and what’s hype.

I'll check it out, but I haven't yet, and I'm not overly invested in certain AI narratives. Nonetheless, this is what people want. I can imagine a variety of lessons that the authors think that we should be learning. However, the outsourcing of search to a chat bot is a good example of where that friction resides. How much of what we do is a function of past limitations that no longer hold in the current moment. How much of what we do is unintentional elements of our activity path? For example, KVR interaction paths naturally follow the currently most popular posts. So, we have to ask what is the utility of KVR in that moment for those posters, vs what is the utility for others, either interacting with only subcommunities, or looking for facts. We also have to ask how much organizations like this even exist to serve a true need if that need is primarily information. Then, if that need for information can be outsourced to a model, even one that finds it on KVR, what need does KVR have for human attention?

Attention is what AI will take from you. Attention from other humans listening to your tracks, or even just reading your words, because, humans can outsource functional intention to agents. .This is happening, this will continue to happen, and it is not stoppable.

[zerocrossing]

I know I keep bringing this up, but you all need to listen to the Shell Game podcast. Season 2. It’s all about a guy who makes a company out of all AI agents and himself. Their product ends up being an AI driven application that surfs the web for you, leaving you extra time to not have to deal with that. “Sloth Surf.” The agents even named it. It’s hilarious and indicative of where the technology actually is, and what’s hype.

I'll check it out, but I haven't yet, and I'm not overly invested in certain AI narratives. Nonetheless, this is what people want. I can imagine a variety of lessons that the authors think that we should be learning. However, the outsourcing of search to a chat bot is a good example of where that friction resides. How much of what we do is a function of past limitations that no longer hold in the current moment. How much of what we do is unintentional elements of our activity path? For example, KVR interaction paths naturally follow the currently most popular posts. So, we have to ask what is the utility of KVR in that moment for those posters, vs what is the utility for others, either interacting with only subcommunities, or looking for facts. We also have to ask how much organizations like this even exist to serve a true need if that need is primarily information. Then, if that need for information can be outsourced to a model, even one that finds it on KVR, what need does KVR have for human attention?

Attention is what AI will take from you. Attention from other humans listening to your tracks, or even just reading your words, because, humans can outsource functional intention to agents. .This is happening, this will continue to happen, and it is not stoppable.

I disagree. I think, for me at least, KVR and Gearspace fulfill two things that AI can't. First off, it's like a cocktail party full of music nerds that never ends, and nerds are argumentative by nature. Not just chit-chat, but disagreements and sometimes fights. AI is always an ass kisser, by design. I don't want to chit-chat with something that is always going to yes me. Conversations are best when there is some friction, which we have in abundance. For instance, this entire thread is a troll. I don't think that's clear to everyone, but it's pretty hilarious to me. Could AI sarcastically troll as well? I don't think so.

Second, AI doesn't have an opinion. I have an internal "list" of forum members who I know well enough to know that if they like something, there's a good chance I will like it as well. I know we share certain opinions. Of course this is never 100%, and that's what makes the world go round, as they say. For instance, the Jupiter 8 is both an amazing synth and an uncontested classic... and also something I'm not really that into. The Grateful Dead is an amazing band full of great musicians... and something that I do not like at all. I don't think AI can have opinions like that.

Lastly, it's a source of information, which AI can be as well, though both can be wildly off. See my second point. I know who to listen to because their information is generally very good.

[ghettosynth]

Attention is what AI will take from you. Attention from other humans listening to your tracks, or even just reading your words, because, humans can outsource functional intention to agents. .This is happening, this will continue to happen, and it is not stoppable.

I disagree. I think, for me at least, KVR and Gearspace fulfill two things that AI can't. First off, it's like a cocktail party full of music nerds that never ends, and nerds are argumentative by nature. Not just chit-chat, but disagreements and sometimes fights. AI is always an ass kisser, by design. I don't want to chit-chat with something that is always going to yes me.

Sure, but, you are you, you are not the sea of people who post rarely. That is, you are much like the music practitioner in my conversation. Now, with respect to arguments, or disagreement, that to me is a function of a lot of factors. AI doesn't always yes you, moreover, there are really excellent ways to anchor AI output. You may find that other people who provide some of that information may not feel the same as you.

However, more to my point, I wasn't saying that my questions were inevitable, I was challenging you to ask yourself what you get from it and also what KVR gets from it.

Conversations are best when there is some friction, which we have in abundance.
For instance, this entire thread is a troll. I don't think that's clear to everyone, but it's pretty hilarious to me. Could AI sarcastically troll as well? I don't think so.

How much friction? How useful is this conversation? Does its intent on trolling, which happens in different registers mitigate its usefulness. However, coming here for entertainment in and of itself is not coming for information.

Second, AI doesn't have an opinion.

Sure it does. However, that's not a refutation of my point, moreover, in this context you mean that it doesn't render subjective judgement on things that you want it to. In reality however, subjective judgment is just objective evaluation under constraint.

But, again, this isn't about people for whom forums provide a social value. The question is can the afford to exist when that is the only thing remaining. The question is will that be the social mechanism that the majority wants as it devolves into just the hardcore.

My final point isn't really open to debate. AI is already stealing attention, that's what it does. That's what you are all complaining about. You're complaining that people are listening to AI music, that's attention that you want and AI is taking it from people. AI is taking attention from google. AI is taking attention from grammar checker products. Anything where artifact creation is integral and difficult, AI is stealing attention. That is not a debatable premise, that is what is happening.


But. But, but. The important part is that we are moving into this and not paying attention to the constraints that are changing and then expecting things to say the same. They won't. People will listen to AI music. IT's already happening. People will ask AI question that they might come here to answer. I'm sure that you've participated in "what's the best reverb" style conversations. They aren't sources of information, they are sources of, often ill-formed, opinion. The signal is buried in the noise, you can often pick it out, but, AI answering that question, even incorrectly, changes things. It's easy to elicit opinion as summary of expert opinion from AI.

Give five good, currently available, reverb plugins for music production. Search online and focus on preferences by professionals for realistic spaces.
Thought for 20s

Audio Ease Altiverb 8 (convolution, real spaces)
Altiverb is the “go-to” when you want actual rooms, halls, chambers, and sampled classic units via impulse responses, with a gigantic IR library and strong search/browse workflow. It’s a staple in pro mixing and post precisely because it starts from real captures.

LiquidSonics Seventh Heaven Professional (Bricasti M7 recreation, realistic depth)
If you’re chasing that “expensive, believable” hardware space, Seventh Heaven Pro is built specifically around reproducing the Bricasti M7 sound. Sound On Sound calls it a strong alternative to owning the hardware, with practical control and great sonics.

LiquidSonics Cinematic Rooms Professional (room simulation that stays convincing)
Cinematic Rooms is aimed at score/post workflows but translates extremely well to music when you want early reflections and room geometry to feel “physical” rather than wash. SOS praises the sound quality and especially the early reflections, and LiquidSonics even highlights endorsement from composer Hans Zimmer on the product page.

iZotope Exponential Audio Stratus (natural/transparent “put it in a space” reverb)
Stratus is positioned by iZotope as the most modern version of their “natural reverb” engine (vs their more “character” line), and it’s widely used in post where realism and clarity are non-negotiable.

Acon Digital Verberate 2 (natural tails, realistic movement without cheesy modulation)
Verberate 2’s Vivid Hall algorithm is explicitly designed to model subtle real-world variation (air movement, audience shifts) while avoiding the obvious artifacts that can make reverbs feel synthetic, which is exactly what you want for believable spaces.

If you tell me what you’re mainly reverberating (vocals, drums, orchestral, synths) and whether you prefer convolution realism or algorithmic tweakability, I’ll rank these for your exact use case.

Note: You should certainly expect these kinds of queries to be manipulated by monetization of services.

[VitaminD]

Maybe EMPs will send us back technologically to the 19th century again.. at least for awhile.

[Bombadil]

Tease.

[ghettosynth]

Maybe EMPs will send us back technologically to the 19th century again.. at least for awhile.

Sure. That won't create devastating consequences overnight in every other area of nearly everyone's life.