▲ 102 points
back
45 comments
The human performance at the bottom of the page was nice. But you can also give a great musician a zipper and paper bag and they can make it sound nice.
It’s just like LLM generated prose to me. The orchestration and motif development is musical sounding, but there’s these artifacts that are like the “It’s important to consider” or “Would you like further suggestions” of the aural world.
I think it would be fun to write up a detailed analysis of a full piece.
It's literally a list of cliches glued together. Some of them are style-defining cliches, but there is no hint of a strong original voice. And it's the individual distinctive voice of each composer that defines classical music. The musical wrapping is the means to that end, not the end itself.
The waltz sounds like elevator Chopin with all the bittersweet irony sanded off (and couple of questionable harmonies.)
It's an achievement to get this far, but it's still very uncanny valley, and (IMO) a fundamental limitation of LLMs that can't see beyond the superficialities of weightings into what the structures mean.
I was actually quite impressed by the non-human version of the Waltz before seeing the video with the musician. It's a lot like the fantastic Succession theme, it's not a generic waltz, it's a bit dark in an interesting way.
I am really interested in this. The current paradigm from e.g. Suno is an all or nothing finished product. Producing intermediate assets allows you to do simple things like proper mastering or swapping instruments or editing melodies etc.
Yeah. I want a DAW with autopilot features (assistance), not for the LLM to wholesale take away my creative input (vibe composing).
This is our goal with https://parture.org!
Is your goal to break down audio to individual instrument's sheet music? That'd be nothing short of amazing.
I agree that what ai music needs to become an industry tool is the ability to create, access and remix parts, but I think tools like Suno have more of the right idea vs tools like this. In order to be able to write intermediate parts properly, you need to be able to understand the whole and what things should sound like when put together, or when the notes are actually played by a musician. Then it’s easier to work back from there, split your tracks apart into stems, transcribe your stems into MIDI, etc.
Suno et al are moving in this direction but I honestly think development will be somewhat stunted until we get a good open source model(s), and something like control-nets.
An impressive achievement indeed. That's exactly what one'd expect from LLM.
If someone wants to see/hear what real music is, listen to this, just to get a reference point. This will make you cry. Literally.
For anyone else tempted to give in to their baser instincts, don't waste your time on the links, it's two very short piano performances. They're good, but it's statistically extremely unlikely that you will shed even a single tear.
I am generally in favour of throwing shade on LLM "creativity", at least thus far, but this is an uncomfortably strained flex. Literally.
I think the "literally cry" thing is a zoomer thing. I watch a lot of YouTube music video channels and the "this made me cry" thing is a meme that shows up with extreme frequency. I don't think we should take it any more seriously than we do "laugh out loud" when the actual response is usually more like a quick exhale.
What is it about classical and jazz fans that makes them so pretentious?
Please listen to the above pieces once a day (it will take only 4 minutes) for 10 days. Then, while speaking about music, you will sound pretensious to your friends. Please try.
(I can send more. This will include jazz certainly. And Beatles... and others. You have to put some effort into it, it will pay off).
I've been studying music for 10 years, I'm not some 12 year old who's only ever heard pop.
Classical music isn't special superior form of art, and people who try to paint it as such tend to be elitists who want to have something to hold over other people to make them feel superior. Kind of like how people latch on to wine to try and seem sophisticated (ironically, wine snobbery has its origin with kings and chieftains of conquered territories trying to fit in with their Roman oppressors, but at least they were cognizant of why they were doing it, unlike modern wine snobs). There's plenty of metal that's far more technical and complex than typical classical music, but you don't see metalheads going around telling people to listen to "real" music.
What's special about classical music is merely that it's a literate form of music. It has an underlying "text" (the score) that everyone agrees on, which represents a somewhat abstracted 'blueprint' of the overall work to be performed. A metal piece doesn't really have that: you can transcribe it after the fact to score notation or tablature, but the result is merely one listener's opinion of how that piece "goes". Jazz music has its "lead sheets" but these are intentionally simplified and/or otherwise changed wrt. the source material taken from the "Great American Songbook" rep (which is far closer to the classical tradition than any kind of modern "pop").
This means that classical music, more than other traditions, is a natural target for both broader academic study as well as automated generation by AI's trained on some sort of existing repertoire.
I didn't study classical, but I was watching some YouTube channel by someone with graduate degree in classical. He offhand mentioned that many of the "compositions" of guys like Chopin were actually improvisations that were later transcribed. I've heard similar tales of Bach, especially his ability to improvise complex fugues on the spot as a sort of party trick.
I just mean to say, I don't think there is a clear divide between a classical mentality for composition and a modern mentality. It is just we study classical hundreds of years after the fact and that academization of the music has lent a particular view to it. If we study metal in a hundred years the same way we study classical (or jazz) it may seem just as rigid.
You're right that improvisation was historically very relevant in classical music and that this tradition was mostly lost starting somewhere in the mid-to-late 19th century as the view on how pieces should be performed became a lot more rigid. It still survives in places, such as among organists, and there are many attempts to revive it. But the fact that it is a matter of academic scholarship and study is not that closely related: there are lots of period-contemporary treatises and 'method' books that discuss exactly how pieces should be improvised and/or performed, often in great detail and depth. You can't possibly have that unless people are very much familiar with the practice of writing their music down on paper. That's what 'literate' means basically, it really is as simple as that. It's also something that other music traditions tend to not focus on to anything near the same extent.
I’m not sure how useful that framing is because improvisation is effectively composition in real time. By writing down an impromptu or something that was played on the spot it becomes composition when it’s crystallized in this way.
Chopin predates audio recording, so it’s not as if someone recorded his performance and then transcribed it like people do today with jazz solos. I’m assuming that Chopin was the one to transcribe and publish after he had a good idea because few would have the recall to do so accurately.
Sometimes compositions are generated spontaneously by the composer (one pass) and sometimes they require extensive labor and refinement.
It’s not to say that classical music doesn’t also include improvisation, and I agree that there’s not a clear divide between classical mentality and modern sensibilties. Although today, the composition (if not scored) ends up taking it’s final form as an audio recording. More and more the composition exists untranscribed in the DAW session.
Completely agree with you, except
> you don't see metalheads going around telling people to listen to "real" music.
Have you _met_ a metalhead? :) saying that as one myself (less now, used to play in bands and stuff), this is one of the snobbiest group music-wise!
> Have you _met_ a metalhead? :) saying that as one myself (less now, used to play in bands and stuff), this is one of the snobbiest group music-wise!
Like when we were 15, yeah sure, everything sucked, Opeth rules, Mayhem is tru Black Metal! etc. Though later on only my really immature friends kept those opinions. But I think this goes for any niche subculture that people use to define themselves. I too am an old metalhead, still practicing! Dress like a 15 year old, crank the volume and bang your fucking head!!! \m/(>.<)\m/ ;-)
I see metalheads casting shade on subgenres of metal they're not into and overhyped bands moreso than entire other genres of music. I remember a lot of conversations about how Cradle of Filth was poseur black metal for high school kids back in the day, for example.
People can absolutely be snobs in other genres, it's a super common thing
I was expecting another song, which is more likely to elicit an emotional response.
Hello fellow emotional music fan. This makes me cry, literally:
The pop composition was nice, if somewhat cliched. (And of course the MIDI instrumentation does it no favors). The guitar solo in the middle was fun - it kind of sounded like it reverted to some of the baroque pieces in its pretraining. Reminded me of the harpischord solo in the Beatles' "In my life".
This is where AI becomes a tool for the artists.
We need an AI DAW, not Suno.
Suno is for people who would otherwise not make music. The grandchild who wants to send grandma a funny song. The marketer who wants a quick solution for a social post. There are valid reasons for both types of AI products.
There's a bunch of AI DAWs out there, or AI plugins/VSTs to existing DAWs. They're not popular because the mainstream consumer wants Suno.
Can you name a few? In the context of this model I assume you mean some kind of composition plugin that will plot notes for you?
A bit off topic, but is there any good recent Image to midi/XML/Sibelius stuff out there?
Soundslice has a fairly new feature for scanning PDF or image into a music score, with export to MusicXML and MIDI.
Generative music is boring not cause it’s necessarily bad sounding music but because there’s no vision behind it, they took in old music, garbled it around in their magic music machine and made new music based on that old shit.
As long as there’s music made by real people this AI music stuff will only be interesting to the very uninspired, the people that will hype up anything that’s AI and commercial projects who don’t have to pay royalties for the music in their advertisements, elevator, phone waiting music and whatever else they need a quick tune for.
Don’t get me wrong, the technology is nice but I don’t see myself consciously enjoying AI music, at least not till I start dementing or smth.
Even the cherry picked examples are boring milquetoast? Maybe this is an example of "this is worst version of this technology" but I'm still skeptical of any generative system that isn't publicly available for testing.
The major issue I can foresee going forward is that the only public domain sheet music available is in the classical idiom. And people don't really care about writing new pieces in a Baroque style. It's a fun exercise, but not really relevant to modern music. The "pop music" listed in the article sounds like "Baroque-pop" presumably because that's what is in the training data. That's why a DAW style system would be more useful and welcome, although there again - what data do you train on? There aren't just millions of publicly available "EDM DAW files" laying around to train on.
> The major issue I can foresee going forward is that the only public domain sheet music available is in the classical idiom.
I'm not sure that this is accurate. You can take a look at IMSLP and see how much music there is that's in a more "popular" style (while still not anywhere close to modern pop, of course). Especially vocal music (most commonly songs) and instrumental dance pieces, which were the most common choices among 19th century amateur performers playing music in their home salons and parlors. It takes a lot of work to accurately transcribe the music from score notation into a form that AI can be trained on, but that work too is happening via e.g. the "OpenScore" effort. So I think we can expect better AI-generated notated music in the future. It could become a nice "toy" domain for sequence modeling (including language modeling) architectures, where high-quality results might be achievable with very little training effort. A bit like MINST is today for image recognition.
Is this an AI generated response? Spending 60 seconds on Google reveals every argument to be nonsense. A simple Google search for "OpenScore" reveals a MuseScore page with only 14 scores (all of which already exist in IMLSP). So what kind of "work" you think is happening there...I haven't a clue. Searching IMLSP for popular songs reveals mostly recordings and badly photo-copied PDF scores - nothing usable in the MIDI, XML, or ABC formats.
I follow a YouTube channel Gamma1734 [1] where the creator who runs it plays a bunch of piano sheet music from the past, a lot of it in the Romantic style. The artists played on the channel are all obscure to me. None of the names I consider "masters" from that period or style. And to be frank, the average quality of the compositions is not great.
My own opinion, as an amateur musician with no classical training, is the music I heard on this AI generated page would not have stood out to me compared to the rest of the pieces on that channel. I would wager that in a randomized test of these average middle of the road composers vs. an AI score generated by NoteGen that almost no average person would be able to distinguish them.
That leaves two categories. The highly trained individuals (e.g. professional composers or well educated amateurs) who may be able to tell a difference (I would love a real study on this and not just "vibes"). And the master composers (Beethoven, Chopin, Brahms, Liszt, etc.) which stand above the average human composers of their own times.
I would admit the pieces composed by the AI aren't at the level of the master, but they do feel indistinguishable to me from the average.
This is notably better than previous attempts. I hear the basic structures that make a piece of music more than a series of notes.
None of the pieces I heard were good, but I think they'd get a passing grade in a composition class.
Here is a great online service using NotaGen where you can directly listen to the pieces and follow the scores on screen: https://piano.fm/
Notes in the left and right hand that occur on the same beat don't line up properly.