A survey published last week suggested 97% of respondents could not spot an AI-generated song. But there are some telltale signs - if you know where to look.
Here’s a quick guide …
-
No live performances or social media presence
-
‘A mashup of rock hits in a blender’
A song with a formulaic feel - sweet but without much substance or emotional weight - can be a sign of AI, says the musician and technology speaker, as well as vocals that feel breathless.
- ‘AI hasn’t felt heartbreak yet’
“AI hasn’t felt heartbreak yet… It knows patterns,” he explains. “What makes music human is not just sound but the stories behind it.”
- Steps toward transparency
In January, the streaming platform Deezer launched an AI detection tool, followed this summer by a system which tags AI-generated music.
- dumbass nonsensical lyrics
- bland basic bitch tone
- superfluous background music
- digital voice that sounds like it’s been through a syth incorrectly
Do you know why I listen to real artists? Because there is a story behind it, someone had an inspiration and wrote a fucking good album. AI has no story to tell, didn’t broke up with someone to write a song, and certainly won’t make star gaze the live concert I went to.
My first experience with AI music was when I was on my usual 90s hip-hop/rap vibe and got recommended some channels with alleged underground hits. There definitely were a couple channels that put out legit mixes that did have a lot of music and artists I didn’t know prior, but one of the mixes was weird. I could tell immediately, less than a minute in, mainly because of the vocals that sounded super generic as well kind of robotic in addition to a very out of place beat that doesn’t sound at all like it’d belong in the 90s/2000s era of rap music. Had it not been for the vocals in tandem with the mismatched beat (obviously created by someone who doesn’t know jack about the music genre and the ear it’s supposed to represent), I might not have spotted the AI involved.
The scary and sad part is that I doubt YouTube will do anything about it despite reports and that there are so many people that either don’t care or don’t know/realise. Only saw like one or two other comments calling out that mix having been made with AI
I don’t listen to anything made after about 2010. I don’t have these worries.
I only watch YouTube videos from pre 2010. You probably wouldn’t know them.
Not sure but I generated this and it’s something I really like https://www.youtube.com/watch?v=XPYk3zjDY-o
I was looking for videogame remixes one day and found a channel doing Little Nemo from the NES. I used to love that game and thought it was an odd pick for remixes, one you don’t see too often so I clicked on it and … it was incredibly underwhelming. I listened for a few minutes and something was kind of off but I couldn’t quite put my finger on it. It was AI of course.
I’m not much of a music person, I’ve been listening to it daily for my entire life but I don’t know much about theory. Still, when it comes to remixes, you can usually tell why someone remixed a song. They like that particular song, or there’s a motif that really struck them. They’ll pick out certain sounds or elements and build on them, single them out and rearrange them. It’s very intentional and you can tell.
AI-generated remixes lack this intentionality. It was like someone had twisted a dial that just said “complexity” and that was it. There were more intricate layers of beats and instrumentation on top, but it wasn’t doing anything. I sat there and listened for 15 minutes and it was like I heard nothing. Nothing new stuck in my head, there was no riff or little melody that made go, “Aw fuck yeah! This is what it’s about!”
That’s how you can tell AI generated music.
Sadly, a lot of slower and minimalist genres have been decimated by it though. Vaporwave, chillcore, dungeonsynth. A lot of these had large bodies of work to train on and it’s a lot harder to tell due to their subtler nature, but you’ll usually notice the artist has a new hour-long upload every day. If you click through it at random, you’ll begin to notice that while the tones shift, the overall pattern of the entire hour-long mix is still kind of the same?
It’s bleak, man. Fuck that shit.
Tip 1: It’s being promoted. The music industry would love to get rid of musicians.
And the Records Department, after all, was itself only a single branch of the Ministry of Truth, whose primary job was not to reconstruct the past but to supply the citizens of Oceania with newspapers, films, textbooks, telescreen Free eBooks at Planet eBook.com programmes, plays, novels—with every conceivable kind of information, instruction, or entertainment, from a statue to a slogan, from a lyric poem to a biological treatise, and from a child’s spelling-book to a Newspeak dictionary. And the Ministry had not only to supply the multifarious needs of the party, but also to repeat the whole operation at a lower level for the benefit of the proletariat. There was a whole chain of separate departments dealing with proletarian lit- erature, music, drama, and entertainment generally. Here were produced rubbishy newspapers containing almost nothing except sport, crime and astrology, sensational five-cent novelettes, films oozing with sex, and sentimen- tal songs which were composed entirely by mechanical means on a special kind of kaleidoscope known as a ver- sificator. (George Orwell, Nineteen Eighty-Tour)
In case you were curious, apparently you can get Free eBooks at Planet eBook.com.
Lol this is how pop music is created with formulaic, focus group approved garbage over engineered to be the most palatable and sell well.
If that gets killed off along with AI-generated music, that seems to be to be win/win.
I genuinely think AI music is better than the shit they play on radio.
“Dog shit is better than cat shit”
By a very thin margin, but yes. I’d rather listen to AI slop than the Human slop they try to pass for music these days.
Edit: Ah shit, I realized though that we are just gonna get talentless pop stars using AI instead. So both dog AND cat shit at the same time.
Brave new world.
literally 1984
But also there is still lots of good human music being created, I guess it’s passing under your radar. Ask around, browse Bandcamp, listen to radio (hint: it’s not just local now, check out Radio Garden).
And here I’ve been thinking there’s been a big wave of great bands lately. Maybe you just listen to crap.
The only thing more formulaic is Bach, who wrote music mathematically.
Hahah he didn’t exactly write according to a formula. That’s like saying “yeah jimi hendrix is pretty formulaic, because he just plays guitar, with a limited number of chords and strings.”
Edit: On second thought, the above is not a fair comparison at all, and there is a point to the mathematical nature of Bach, I just couldn’t express it in a coherent and snarky way at the same time.
I have yet to hear an AI generated song that didn’t have some obvious tells in the vocals. Like how you can hear autotune, but it’s even worse than autotune. Crackly/crunchy, heavily distorted but only on certain words. Weird pronunciation or annunciation.
You have yet to notice an AI-generated song that didn’t have some obvious tells.
https://www.youtube.com/watch?v=a515IaGVSmE&list=RDMMa515IaGVSmE&start_radio=1
I thought this was real
I fell hard for Dysmn, who was suspiciously dropping new music every few days. I really liked the sound, and I haven’t found anything that sounds like that since. 270+ videos in under 2 years. I realized it wasn’t human after a month or two.
Soooo, if anybody knows a great jazzy EDM metal noise, let me know.
I looked that up, and it actually slaps. Not really my genre, but I can see how people would assume it’s real people.
Humans are primarily visual creatures, so we can detect the slop in AI images a LOT faster than we can in audio.
Human artists are going to have to get a lot weirder to out-innovate AI music, and I’m actually happy about that. Weird music is the best.
I think this might be a way AI and humans can actually work together. A lot of musicians are creating digitally anyway. They might have to rely on drum loops if they aren’t good at beats. Iterating with an AI to get the right beat would be better than a loop, and still be a human making it.
However, I have reservations about the AI doing everything with little or no human involvement.
Iterating with an AI to get the right beat would be better than a loop
That’s just like, your opinion
Right, pretty sure everyone but you knows it’s an opinion.
I think (think means opinion) most people, besides you, would agree that having a beat that is unique and not just downloaded from a loop library will produce a song that is more original.
Do you create your own samples from scratch too? Kudos if you do, I haven’t done that since the 90’s, and I consider my music original. Ever used an arpeggiator? Or randomized patterns? Used a synth lead that came with the synth?
The distinction between automation and AI breaks down somewhere.
I like to take real life sounds and put layers of effects on them to make them indistinguishable from the original. My drum is me playing my cat’s butt (meow, slap). I love to take quick sounds and slow them down ~100k% and get really weird atmospheric drones, etc.
When it comes to involving AI, which I haven’t done yet, I’d love to use it to quickly iterate over ideas. I spend a lot of time chasing an idea and having it fall apart. I like the process, sure, but I’d prefer to have more wins.
“In my imagination I speak for all people and we all think I’m right” chatGPT is literally rotting your brain bro
“The main issue musicians are struggling with is originality; AI can help :)” get a grip. Notice how in order for you to insert “ai” into this process you have to construct a fantasy where a totally uncreative person lazily drags a low quality beat into their project without a further care in the world? The root of your little fantasy here is simple and inescapable: AI is only appealing to boring, uncreative, soulless people. I know it, and you know it, which is why out of all the limitless possibilities your first argument is that it would be superior to something you already consider shit.
Strawman much?
“I refuse to have a polite conversation” - you. It’s exactly what you said, because it’s in quotes and I know how to use quotes. Hah.
I disagree.
If a person plays guitar-but not piano, and they feed their track into an AI and ask it to generate a backup piano track with specific instructions about how it should sound, they should be… denied that?
What should they do if they want a track that involves an instrument they can’t play? Hire someone? With what money? lol.
The artist is upfront that it is just one person making music, and they say it is AI assisted in their discord.
Can’t say any legit bands that sound exactly like that, But some of the guitar and metal notes sounds inspired by Polyphia, Playing God (sorry for the yt link) might scratch your itch, otherwise look up the math rock genre, might find some gems there. Wish you the best of luck!
Gonna make a quick edit, Unprocessed 100% deserves a recommendation in this genre. Occasionally have some EDM but mainly more on the metal side, but still have some extraordinary strings akin to Polyphia
I’m enjoying a lot of Unprocessed’s Angel album. I’m also going to have to look at Polyphia and Playing God.
Math rock is one of the ways they advertised themselves. Also djent, but I’m not seeing that.
Check out a band called Unprocessed. I just found them after they did a collab with Polyphia’s guitarist, another band to check out. It really sounds like this style.
You are the second person to recommend Unprocessed. Almost 10 minutes into the Angel album. It’s pretty good, definitely hits the target.
I’ll have to check out Polyphia as well.
This description of AI songs could be a lament about most pop music: formulaic, sweet, generic, produced in a studio to sound perfect, not human. Works on radio or Spotify, but not so much for a live audience.
Sure, that’s hard to detect. AI reproduces what we’ve been exposed to for decades.
The drummer sounds like he has to many arms.
And the guitarist and the keyboard players sound like they clearly have more than 5 fingers on each hand. 😋Drummers very often sound like they have too many arms. Multiple percussion tracks can be layered, and also, that’s kind of the point of trickier drum beats.
That’s ridiculous! Drummers can’t have too many arms!
I told the AI to make a drum track with a one armed drummer and I got copyright striked by the Rick Allen Estate.
every progressive genre wants to speak with you.
So Phil Collins and Mike Rutherford? /s
davie504 be like
Could just be Count Rugen though.
Rush?
That’s blasphemy!!!
I’m just saying. I hear about a drummer with 6 hands, it’s gonna sound like Neil Pert. Dude is part octopus.
You’re right. I miss that guy a lot. Every once in a while I’ll watch a ton of his videos in remembrance of his crazy talent.
I’ve been trying to figure out if Stone Rebel is an AI band or not. They started in 2018 and have put out something like 77 albums since, but it’s relatively simple instrumental. They have almost no information online except a claim that they’re “based in France”
Honestly can’t tell if they’re a legit yet very private group, or if they were early adopters of procgen music
all the album art is ai and there’s no mention of them touring or anything. Bandcamp also allows ai artists, and that’s where I found their album art
77 albums since 2018?
It’s AI. No question.
Sometimes experimental/improv groups can basically make an album a day if they put their mind to it. Often live recorded with minimal post processing. It’s far from mainstream but it can be surprisingly popular among the right audiences.
Haven’t checked out this particular project but it’s possible
AI wasn’t making anything close to listenable music in 2018. If they have songs from before 2023 they’re human.
I’m in the camp of, “if it’s good, why should I care?” However, I’m all for transparency! Passing off AI-generated music as human-generated is fraud. Be honest!
There’s a LOT of grey areas though. If you’re a vocalist and you’re using an AI-generated background? How’s that any different from pressing “play” on a sequencer or even an audio file (of some sequenced or drum track)?
If you’re a lyricist, the actual music isn’t as important as the lyrics. Does it matter if they used AI to generate the music or should every lyricist be forced to pay someone to make the music for them or master an instrument (or sequencer)?
What if you’re trying to translate your music into a different language and use AI to translate it? Is that AI-generated music? You can give your whole damned song to AI and it’ll convert to a different language in-place without having to re-record it. It even uses your singer’s voice!
To me, it’s incredible technology and it’s enabling artists of all kinds to do cool things with their music. It seems rather paternalistic to suggest someone’s creativity doesn’t “count” if they didn’t sweat or spend years practicing to create it.
You must consider that the AI “helping” the artist is built from the stolen work of countless artists. Regardless of use case, the tool only exists due to theft. Plus, this tool exists as a way to not pay talent for content.
Since the bread and circuses machine must keep dispensing to keep the masses anaesthetized, the elites need a way to cut the costs or they will lose points are their net worth scorecard and get made fun of by the other billionaires.
Not to mention, AI is a shortcut that does not generate skills besides prompt engineering. We have research proving this with students and the labor force losing reasoning and straight memory by handing off to “AI”. Part of being a musician is the effort and practice and knowing an instrument. Asking the clanker for a tune because learning takes too long or is too difficult goes along with what the article says for detecting it. The work will be emotionless and have no soul. Musicians are allowed to make choices for their music, of course. AI rounding out an artist’s tools is what it is. I view the tool as a corrupting force but, it’s their perogative. But people without no knowledge or skill for making music cranking out these generic sounding similacra to make money is always going to set my teeth on edge.
Edit: spelling and tense correction. Revision and expansion of idea to express less derision.
You must consider that the AI “helping” the artist is built from the stolen work of countless artists.
I feel like this is conflating two separate arguments.
Is AI music good, versus is AI music moral.
If AI removes the skill and expertise required to perform a task, it’s fucking trash
If a vocalist can’t play music, they should find people that can, and work with them. If they use AI, they’re are trash.
Bullshit. This only applies to fully prompt generated AI music. Tracks that heavily rely on AI based tech as a part of the process are harder to catch, and tracks that only use AI for mixing and mastering are impossible to detect.
I made a track but used AI to autotune and morph my voice to that of a woman’s. It even allowed me to tweak the expressiveness of the voice. The track is 95% human made but the vocals are AI modified. I’m willing to bet that the ration of AI use in a lot of pop music and EDM is a lot higher.
PS: I make music for myself, as a hobby. I just wanted to make something to share with my friends. If you want real music, try bands like Wet Leg, IDLES, GEESE, etc who lean into making low tech music.
EDIT: Thia is an example of a song that is fully generated by AI. All that was fed to the prompt were the lyrics. The AI did everything else itself, including picking the genre. I shared it with a few people to see who’d figure out it was AI slop.
Compression throws me off.
That AI ‘hiss’ is ringing it my ears, but its very, very similar to YT uploads that have been re-encoded like 10 times. Which is a lot of them.
Out of curiosity, I made spectrograms of the AI song and concert that sounds ‘clean’ yet kinda noisy/compressed to me. I won’t tell you which one is which:
Source song, if you are curious: https://www.youtube.com/watch?v=MydFq0io-tQ
That song was a mistake that got generated when I pasted the wrong clipboard into Suno’s Lyrics window on my phone and accidentally submitted it. It has not been processed at all. Suno has a “Remaster” feature that when you are happy with the generated song you can give it a few automated “mix and mastering” passes to generate a cleaner and more dynamic sound.
I’ve mainly used Suno to mess about so I didn’t want to pay for the upgrade.
If there’s ever a need, you could try whatever open generation models use for post processing.
