This is literally some of the impossible sci-fi tech I dreamt of as an undergrad. Crazy. I'm still a bit in disbelief how fast things currently move on this front.

Interestingly, suno.ai is also able to imitate the very robotic and staccato-like intonation of Vocaloids: https://app.suno.ai/song/f43e9c46-92d3-4171-bdd9-026213d6772... - everything comes around. :)

* sublicence - "sublissence"

* fitness - "fisted"

* infringement - "infring-ment"

* liable - "liar-ful"

It's also obviously not a pure human voice recording as the pitch transitions sound heavily auto-tuned or electrified (think Cher's "Believe").

I anticipate people becoming experts in detecting AI-generated vocalists in much the same way that we can currently detect AI-generated images due to abnormalities especially in details like ears or fingers.

I am certain there exist weird wines that could fool me (I’ve had a few really weird wines) but typical shit from the grocery store, I’m gonna be able to tell at least that much. I might even ID them more precisely than red or white. It’s not exactly subtle…

Then again I don’t have a clue how someone could fail to tell which is coke and which Pepsi in the “Pepsi challenge”. They’re wildly different flavors. I can tell by smell alone.

So it's less a case of "they cannot distinguish red from white" and more a case of "they went along with a suggested classification". I feel like this is a weaker result, although it's still a little surprising.

For comparison, imagine someone showing a piece of Picasso to art critics and saying "Could you please describe the artistic significance of this painting by da Vinci?" The critics won't start using terms commonly reserved for Renaissance era; they'll say "What the fuck are you talking about, this isn't da Vinci."

The hard part is identifying the type of wine, but many of my wine-drinking friends can do with ease. We've tried the "test," having me or someone else randomly purchase wines from the closest store and then serving random samples to them while they're blindfolded. They're able to identify the specific variety more than 4/5 of the time.

Source: my wife's godfather did the studies for that[1] two years ago.

[1]: https://www.isvv.u-bordeaux.fr/fr/diplome-universitaire-dapt...

What’s a good place to find out the SOTA of the custom tooling and workflow?

"Quadrupled" is a very specific and quantitative word. What measure are you basing that on?

IMHO many of the successes of "artificial intelligence" come from "natural stupidity". Humans have many glitches in our perceptual mechanisms. The AIs that end up going viral and become commercially viable tend to exploit those perceptual glitches, simply because that's what makes them appeal to people.

I just rewrote the lyrics as "paste-ing" and it sung it perfectly afterwards.

    The cake is a lie, but the music is real

    It’s all fake when the truth is revealed

Or think GLaDOS. Pretty sure that's not a coincidence.

(like it's sort of no difference than paying someone to voice something and share it)

I think the stuff that is completely generated with no human in the loop is a different category for me because it can be used for things at scale like, bots on social media, or ads in a podcast generated just for you, etc. As long as there is still a human in the loop making the editing decisions, it feels not categorically different from the world we have today.

Like, there's been computer-singing voices for awhile, but they always sounded pretty robotic and goofy (e.g. Microsoft Sam), and I think for a long time people just assumed that to get mostly-realistic voices, you need an actual singer. Yes, it still requires a bit of human tweaking to make it perfect, but I suspect that if put to the test it would reduce the cost of making a song substantially.

AI/Human combos can still be valuable. More broadly I'd argue that that's how almost all tech works. E.g. there are still textile workers, just many less of them producing much more clothing.

People fail to identify even the most basic and obvious fakes, but somehow there’s a group of people who think that as fakes become harder to distinguish from reality, we’ll all magically become experts at it. We won’t. People’s ability to detect fakes will get worse, not better, as a consequence of more prevalent and better fakes.

https://www.youtube.com/watch?v=1U1HMqtam90

Free accounts are queued so it depends on load and I don't think the v3 model is available to them.

Counter-point, we've been listening to a rock song about the moon given the words from one of my kids books all morning.

People will 100% listen to (edit - I never finished the sentence) things that sound fun. It might not bring me to tears or stop me in my tracks but lots of things are just fun.

> People aren't going to hang AI paintings in their houses

People will absolutely do this. If AI systems can make nice pictures people will hang them in their houses. And they can make nice pictures.

A lot of people are very confident about this and I dont understand why. The same was said for jazz and comic books. But I am listening to jazz with comic book posters on my wall. There were different reasons to give the same statement, but it almost always turns out to be wrong. Humans like what they like and seldom judge an artwork for its process (outside of a very small niche community).

>Humans like what they like and seldom judge an artwork for its process (outside of a very small niche community).

That's true, but how do you zoom out of process? This is beyond process. I would just say most people don't like inhuman things.

It's an omellete. There is no Dolly Parton behind an AI Jolene or a Michael Jackson turning a 4 track tape into a musical masterpiece. The journey and personalities are what contextualize the sound, without AI that context is gone. That's why I think it will just be used for cafes and things like that where they want to escape licensing fees.

As for consumers - I believe people will see AI music consumption as a way of supporting the new technological powers that be, and the act of listening to human-made music will have an element of counter-culture baked into it. I'm a professional musician and I have a very physical reaction to sound. Once I know it's AI my goosebumps fade.

Another lame incarnation of a tech that will also fade like crypto and everything else. The types of personalities who will leverage this tech are not the same personalities that make the greats.

I'm not worried.

And anyway, is it measurably different from art produced while tripping on LSD or in similar states of altered consciousness, such as schizophrenia, dementia, or even depression, which often produce things many people would not describe as regular?

The future is obviously a form of custom AI Muzak/Ambient music with a few pop stars for people to focus on.

I am a big fan of more art type music and guess what? No one listens to it. My fav album of 2023 has 6.4k views on youtube. At least a 100 of those are mine. No one listens to this stuff. People watch video critic reviews of more art type music than the actual music itself.

Even if the AI music is extremely good, it's just missing the fact that it was made by a person, which changes the experience entirely. I think we're more likely to see musicians and those top10 artists leverage AI without explicitly saying so.

I expect we will have a daft punk moment where someone is using exclusively AI and later unmasks that it was all AI, and as soon as that happens the music is disconnected.

Same with AI art. I can see something and be duped and go "oh wow!!!" and as soon as I know it's AI the caring leaves my body completely and reverence and interest is lost.

It's better than AI, even this incredible mindblowing suno thing. Production value counts.

There’s certainly a point where this synthetic music gets good enough to replace the elevator music Muzak crap that they have to pay $2000 to license.

I need to stem it, fix it up a bit, and remix for stereo in a DAW but it's much better than I expected for my first ever piece of music. Obviously it'd take a lot of work to create a Hans Zimmer level OST from the tool but IMO it wouldn't feel out of place on a Ludovico Einaudi album or on some Spotify or Pandora classical radio.

I don't think musicians and composers are going to disappear as a consequence of this technology, in the same way that theatre actors were not made obsolete by film. What I do think is that a whole new category of professionals will be created - musicians and composers who get paid to train AI models. I bet it will pay better then the laughable amounts that are streaming royalties.

When they allowed longer text inputs, and for faster rapping, I can really see this kinda thing taking off with L1s and med students.

Like the Animaniacs song about the state capitols.

Or like a Homeric epic that is meant for remembering and singing.

The method of loci may have a new competitor as a way to remember things here.

Musically, however, I can't help but notice that these models are still very far from being able to generate something interesting: from harmony, to tempo, to musical structure, to dynamics, everything is muddled and without structure. I guess there is still very much to work on, and I am not sure that purely generative models can attain higher levels. Maybe a mixed rule-based and generative approach would do?

The progress is really fast in this field, I really do not know.

The github page for bark links to a page about chirp, which returns a 404 page for me [2]. My guess is that the model used for suno.ai's song generator isn't too much different than the text to speech model.

I also have a hunch is that it was more like a coincidence than intentional that the bark model was capable of producing music, and that was spun off into this product.

Unfortunately, there seems to still be issues with bark when generating long (like book length) spoken audio. Which is too bad, as someone who's worked jobs that require lots of driving, it would be awesome to be able to have any text read to me in a natural sounding voice.

[1]https://github.com/suno-ai/bark [2] https://www.suno.ai/examples/chirp-v1

I wonder if these models would do something better if the text were poetic or punctuated differently.

[1] Ha, the poor millions of dumb minions who put their work on the web thinking it might be fun for others or garner themselves a small following, they didn't check the terms of the EULA!

"Write a sad song about the MIT license" is certainly such new input, and if I was commissioned to write the song it would be based on inspiration (i.e., "use training on") music I have heard or studied. And yes, none of the musicians I have listened to or have studied will benefit from the endless money fountain I'd acquire from composing such song.

In the case of AI, a person puts in minimal effort to generate something that devalues the work of all the people who did put in effort.

When someone needs something composed, they don't learn how to write music. They pay someone else the bare minimum, e.g. a few bucks on fiverr. The person will spend the least possible amount of effort to try to make their life go around with the little money they got.

When you then use an AI model, the work done for those five bucks is replaced by work done for almost free.

Neither the person you would hire or the AI credited those who created the material they trained on.

Speak for yourself! There is only one thing that scares me more than composing music, and that’s paying somebody a few bucks in fiverr to do it for me.

Although I suppose royalty-free stock music is the norm nowadays for most commercial uses, which takes it a step further, anonymizing the composer entirely...

Worded differently: people who couldn't otherwise produce skill-based works of value have had the barrier of entry lowered for that specific medium of expression, allowing for more works across a wider spectrum of skill.

Flooding the world with unpolished, unpracticed works, AI-tuned to the level of being mediocre, is a creative and intellectual dead end.

This is what tools are.

Cheap digital tablets have done away with the need for expensive consumables. You can just download a different brush style instead of learning a physical technique. No waiting for paint to dry or smudged pencils. The barrier to entry for painting has dropped to a one time investment of like a hundred bucks. Almost nobody mixes their own paint, nor stretches their own canvas. Those skills aren't needed anymore.

It's possible to build very precise machine parts by hand. It's very difficult and requires great skill, so nobody does that. Some do and are admired for it, but everybody else uses precise machines to make precise parts with nearly no effort.

It's just a tool. Only difference is that we had assumed art would never be automatable.

Objectively, I don't think this is a bad thing. It doesn't change the subjective value of art any more than the average cartoonist devalues the Mona Lisa. It's just a new form of art, there will always be people mixing their own paints and stretching their own canvas, just as there always has been.

It's only a problem because in our society you either have a job or you starve. No one can afford to be an artist. Those that do tend to grind out as many pieces as fast as they can so they can pay the goddamn rent. If not for that, these AI tools would be pretty cool.

* That "_any_ form of creative expression" is a viable creative substitute for people wanting to create in a _specific_ medium of creative expression -- especially those that had a high barrier of technical skills required to be seen as "good enough" to share.

* That a person who has an idea for art will put in the necessary time to become proficient enough to create that "good enough" art through traditional means (IMO demonstrably incorrect), and that is preferred over that person just not expressing a lower-quality version of that idea at all.

* That those who use AI primarily want or expect to "perform at the level of the practiced and talented" (i.e. top-tier art) rather than using it to produce art they otherwise couldn't have, even at low- and mid-level qualities.

* That there is no skill or talent in using AI tools to produce art (or that the skill or talent using AI tools is meant to be a full replacement for traditional artistic skills or talents).

FWIW, I'm a long-time sketch artist and acrylics painter (~20 years). There are many mediums, subjects, and styles that I'm not good at -- and I enjoy using AI to express myself in those areas (and have also liked using AI to create songs to show to my more musicially-adept wife...). But even in my own wheelhouse (landscapes and still life), I also often use AI to brainstorm composition, perspective, colors, textures, lighting, etc. It's a great tool for experts to lean on, but an even better tool for non-artists who couldn't or wouldn't otherwise share their art.

Why is this the line? Where are the complaints about people using pianos to achieve rather precise notes instead of using their own voices? They are just untalented at singing and their use of any tool to create sound is of no net gain to anyone.

This person: https://www.youtube.com/watch?v=IbUE-LxhUR8 ? They're recording and playing back on a loop! They should record full repeated playings, any use of the recording is of no net gain because it could be achieved otherwise.

Songwriters? If they write lyrics and someone else sings them the result should be cast into the sea - it's of no net gain to anyone because they did not create the sounds themselves.

Composers? Frankly pointless.

> Flooding the world with unpolished, unpracticed works

I hate to break it to you but there are a vast number of terrible works of art out there already.

> What this tells you is that skill or level of performance is not the barrier, but a means through which great things CAN be achieved (i.e. necessary, but not sufficient)

If it's a necessary thing, of course it's a barrier. That there are two barriers doesn't change that.

That a person can't sound like the weighted average is human limitation (although with modern pop people do get quite close!), not because new singers aren't trying to. That of course adds variation that we appreciate, but doesn't change the underlying similarity in how acquired skill is mimicry of those who acquired it before us - with very rare exceptions.

What next, are we going to argue that what programmers creating new programs are really trying to do is generate a prompt-weighted average of the bytecode of every program they've ever downloaded, and all that business analysis and functional spec and use of high level programming languages and expressed preferences for coding standards is irrelevant?

That's just a bias.

> natural qualities to their voice

That's the physical limitations I referred to, which isn't something humans tend to be happy about but can sometimes end up being a differentiating benefit.

> What next, are we going to argue that what programmers creating new programs are really trying to do is generate a prompt-weighted average of the bytecode of every program they've ever downloaded

That's a horrible strawman. Do you as programmer often read and write bytecode directly?

I'm beginning to assume you're an LLM, because I'm not convinced a human would honestly try to argue that their emotional reaction to their favourite songs is basically equivalent to flipping the values of some bits to ensure that they generate music more similarly to them.

> That's a horrible strawman. Do you as programmer often read and write bytecode directly?

As an improvising guitarist (even a very mediocre one) my creative process is even further removed from an LLM parsing and generating sound files directly....

We are all nothing but a horde of molecular machines. Your "you" is just individual neurons reacting to input in accordance to their current chemical state and active connections. All your experiences, unique personality treats, and creativity you add to the process is solely the result of the current state of your network of neurons in response to a particular input.

But while an LLM is trained once and then has its state fixed in place regardless of input, we "train" continously, and while an LLM might have experience of an inhuman corpus for a certain subject, we have many "irrelevant" experiences to mix things up.

Your "prompt" is also messy, including the current sound of your own heartbeat, the remaining taste in your mouth from your last meal, the feeling of a breeze through your hair as it tickles your neck, while the LLM has just one, maybe two half-assed sentences. This mix of messy experiences and noisy input fuels "creativity". You don't think "I need to copy XYZ", but neither does the AI. You both just react.

In some regards our chaos is better, in others it is worse. But while the machinery of an LLM still does not even remotely approach a brain, we should not forget that we are nothing but more a cluster of small machines, assembled from roughly 750 MB worth of blueprint.

All of the creators and subjects of meme formats... Should they receive royalty every time you post some inane mashup?

Hell, even with viral videos it's relatively common that normal people can share away while entertainment companies and influencers are expected to pay for a license.

With memes it isn't clear exactly who made the first template, and the creation of them doesn't revolve around specific people in the same way, nor are they meaningfully tied to profits.

When creators post their content online to be shared, they do it with the focus being on reaching individuals, not for it to be sucked up by soulless companies to extract all value without the intention of giving back.

The Office, The Matrix, Lord of the Rings, Django Unchained, Game of Thrones, etc

These works have identifiable creators.

We're not talking about any of those things. We're talking about wholesale digestion of the entirety of human knowledge by automated means, which is now not just theoretically possible, but routine.

This is the real problem, right? People don't dislike generative AI, they dislike the attention economy. Yet I see more disgust towards AI than the company policies which suck. I don't understand why.

Stephen King won't be able to remember every word of every story he's ever read. And if he wants to make something "Lovecraftian", it'll be what Stephen King thinks is Lovecraftian. And there will be something to that. Some bit he believes is more or less important than other people And those bits are what makes Stephen King, Stephen King.

Everyone has had access to the same material King read. Access to the same tools he used to create. Everyone had the chance to effectively be Stephen King. But there is just one. Because there is some unique bit of observation or recall or combination of such things that is unique to King.

And from what I've seen so far, these LLMs can't do that. There is a missing element of pure imagination.

Saying they cannot create based off of a vague suggestion is very much in line with that claim. I consider it a vital difference between Stephen King being inspired and LLMs mashing training inputs together.

break up chapters or sections of the college textbook into suno songs instead - itd be maad interesting how much better that wouldve helped my studies. monotone computer voices of 10+ years ago will put you to sleep.

    The software is provided (as is)
    without warranty, of any kind
    express, or implied.

And some of the generated phenomes actually just sound like stylistic auto-tuning. I kind of like it.

I'm sure many have already observed this, but I think the thing that most artists fear from AI is not that AI will be able to produce works on parr or superior to human works, but that most people won't care enough to value the difference.

I'm not sure whether I've just run out of credit, or Suno actually knows what the political sensitivities of the text might be, but I can't generate a second amendment song.

SVD doesn't have well-controllable motion and is utterly blown out of the water by Sora, but it's what we have right now.

Here's the resulting vid, "a death metal song about a macro photographer":

https://www.youtube.com/watch?v=kNVRQ1Zg-a0

If you only want a video file from Suno to share with the default static lyrics screen on it, hit Download Video from the three-dots menu.

I've dabbled in music production and this is just unbelievable.

Both amazing and a bit sad because this is already so much better than would i would have anticipated.

First illustrators, copywriters, then VFX guys, and now music. We're going to loose so many jobs in the creative sector right?

Long-term coherence, reasonable-ish melody, all on top of very unmusical text. Very impressive.

    The cake is a lie, but the music is real

    It’s all fake when the truth is revealed

Anyway, what will be interesting is when this can be done locally on consumer hardware with open-source AI, a nice UI and Vulkan/DirectML GPU inference.

Soon we will have 'preacher's in a box' that will sing to lift you up, mentor you, guide you through life. Most will even be 'non-religious' but will basically become your religion, your guide through life.

Your version unfortunately sounds much more plausible and profitable.

If this is legit, the Spotify spam is going to become atrocious and probably unmanageable.

I had done a folk song version of my resume. It wasn't going to become a hit or anything, so I don't see this replacing any real musicians, but it absolutely worked to create a passable performance as a song.

Current generated songs are made like sentences where you hear entire song without much structure

* Put a poem my late mother wrote to music for her memorial

* versions of 80's new wave songs

and they came out so lovely compared to what I'd be capable of as a musician, but puts me in the role of a "producer" of sorts tuning the sound and vibe. Really well worth the money.