Musicians algorithmically generate melodies, release them to public domain | Hacker News Reader

Musicians algorithmically generate melodies, release them to public domain

vice.com

202 pointsdylan60495 comments

lots of discussion over here on this topic from two weeks ago https://news.ycombinator.com/item?id=22301091

And discussed in 1692, by Bernoulli, Mersenne and Kircher

http://articles.adsabs.harvard.edu//full/1979HisSc..17..258K...

I'd like this to work, and kudos to the creators for having an interesting idea in a problem area. I've been surprised that copyright plaintiffs have been winning these cases, e.g. Marvin Gaye vs Robin Thicke, the Katy Perry case. Maybe this "all-melodies-ever" idea has an off chance to defeat copyright claims based solely upon melody, although I'm doubtful (that said, I haven't thought deeply about it and copyright law is filled with nuance).

But practically, most of these claims are not based solely on melody. For instance, a plaintiff beat Katy Perry based on expert testimony that the songs were similar "with regard to their pitch, rhythm, texture, pattern of repetition, melodic shape, and timbre."[1] The way these cases are being decided, plaintiffs could avoid dismissal by claiming copyright in some nebulous combination of protectable elements, even if melody is a key part of similarity.

One ray of hope: The Ninth Circuit Court of Appeals (which decides most of these cases, since it encompasses LA) is reconsidering these issues in the case where Led Zeppelin was accused of ripping off the Stairway to Heaven riff. The lawyers made their arguments last year, and we are waiting for a decision. A good decision could clean up the law governing these claims.

[1] http://www.fordhamiplj.org/2019/09/27/katy-perry-lost-her-da...

This is a great video about that Katy Perry case by composer Adam Neely. https://www.youtube.com/watch?v=0ytoUuO-qvg

That was amazing, thanks so much!

Arrangements are definitely copyrightable, completely independently of the melodies used to build them.

Isn't it still a derived work though? (Serious question, not rhetorical.)

If their legal argument is legit, it also means they just infringed upon the copyrights of most modern songs by attempting to release their melodies into public domain.

Someone who owns the copyrights to a song should sue them for a paltry sum, just to see how things play out in court.

For there to be infringement, someone needs to prove that copying took place. Because of the algorithm these musicians used to generate this, they can prove mathematically that copying did not take place.

So if an algorithm you wrote produces a melody identical to a copyrighted work, you can claim copyright of the melody too?

I don't believe there's an objective right and wrong answer here, we're still figuring out how our legal system should adapt to modern technologies.

Part of the law is that the infringer has to have had access to the work that he/she is accused of "stealing"

An Algorithm has no access and could have never heard the work in question..

The individual writing the algorithm certainly can have access. If I copy a song, I can't say that the tool I used didn't have access as a defense because that makes no sense.

> The individual writing the algorithm certainly can have access.

Sure he can have access, but looking at the algorithm you can prove without a doubt that the algorithm couldn't have access to it.

> If I copy a song, I can't say that the tool I used didn't have access as a defense because that makes no sense.

You could have given access to the song to the tools that you used though.

I think there is a difference in creating an algo that generates all possible solutions vs one that generates one specific solution.

No, it's not. One that generated a specific solution is given specific parameters. And those parameters can be, in their turn, just plugged in by a loop, hence generating everything.

Usually songs are between 2 and 5 minutes long. Therefore I can, theoretically speaking, write an algorithm, that can use linear interpolation, to generate all sounds, for all octaves, using 44KHz as sampling (the most usual track these days), and you can bet your rear latest pop or rock or whatever song will be generated as well - leaving me with only showing to court "here Your Honor, after looping through, my algorithm at mark xxx gazzilion generated this awesome piece of melody that you know from Youtube as <Lana Del Rey - Blue Jeans>"

So the Library of Babel for music?

https://libraryofbabel.info/bookmark.cgi?eofpgjcg.ggyvpzpti,...

^ search that bookmarked page in the Library of Babel for “jeans” ;)

On another note, since melodies can be encoded into a text format, the Library of Babel has actually done this too.

Heck, there’s a page in the Library of Babel that reads “Your honor, Lady Gaga will one day steal this melody from the library of babel, encoded here in text format: {melody for ‘Shallow’}”

which may look something like this (search “lady gaga” https://libraryofbabel.info/bookmark.cgi?vakuygoqaattmqm315 )

And discussed in 1692, by Bernoulli, Mersenne and Kircher

http://articles.adsabs.harvard.edu//full/1979HisSc..17..258K...

Sounds awesome. I agree with this 100% (both your sarcastic statement AND libraryofbabel)

Sounds a bit like parallel construction...

> For there to be infringement, someone needs to prove that copying took place.

The argument isn't going to help. If you are creating every permutation in the purpose of publishing, you are inherently aware that you are intentionally copying published works (and are familiar with the concepts), even without specifically choosing to reproduce in each individual permutation.

What if a good actor who owned an infringed melody sued them with the intention of losing to set a good precedent?

See vidarh's very good post which I won't repeat.

A court trail is not much a precedent. The judge can decide to not allow you to mention it. Even if a sympathetic judge (meaning faking neutral but really biased) does allow it, it will be with instructions that allow the next jury to ignore it.

To get a precedent you need to go to court. Then you need to appeal, and a court needs to agree to hear/rule on the appeal. Here you have a problem though: even if you conspire to lose, someone else can file an "Amicus brief" https://legal-dictionary.thefreedictionary.com/Amicus+brief and make a compelling argument that you overlooked. A first round appeal may not get this, but by the time it get to the supreme court (probably not even you - it generally takes dozens of different cases before it gets that far) there will be a lot of interest and your attempt to lose can still win because those who want to win can pour enough into their briefs.

Courts and bar associations tends to take a dim view of manufactured disagreements, so they'd better make sure to be good at just seeming incompetent.

But it's also not likely to be very effective unless you manage to trick a judge into issuing an overly broad judgement and manage to get it upheld on appeal.

The legally sanctioned version would be to seek a declarative judgement.

They could, but what's the consequences of the court ruling in their "favor"? - Consequence would be that those artists have copyright to all future music, thus the one who sued couldn't create anything new and claim copyright in it but would have to pay royalties

I considered doing something like this project, but this was too big a red flag for me to ignore. It seems tractable to go through everything in the library of Congress, if it's already digitized, but that took this from pet project to big hustle

Well, maybe it's fair use to pull just the melody out of a full song, but it wouldn't be in the other direction since the original works here are just the melodies.

> And in MIDI format, notes are just numbers.

>"Under copyright law, numbers are facts, and under copyright law, facts either have thin copyright, almost no copyright, or no copyright at all," ...

I’m surprised they went there, because that makes no sense. Nowadays books, movies, TV shows, podcasts, and pretty much every other form of media consist of a stream of numbers, so by that argument nothing would have copyright.

What if I can encode a movie by n^m then it becomes a crime to share that number. copyright makes numbers illegal, it's hard for most to understand since these are very large numbers. what if I can express the information as 2 numbers. (n) and (m) n^m. .... then n and m individually are not illegal, but someone telling you that you can apply ^ to it becomes illegal. or maybe that is not illegal, but if you consume it as the appropriate media it becomes illegal. ... the challenge today is that sharing these numbers or possession of these numbers are a crime. sure, you can call it a file, but it's just a number. if i take a foo.mpeg file and rename it foo.dat or encode it as a zip, gz, string of 1 & 0s, uuencoded data, the law makes all of that illegal. For now these numbers are illegal, but should numbers be illegal?

I read an article (>10 years ago so sadly I don't recall the title) about how computer science people see data differently from law people. Computers know nothing about the intent of data, it is only bits. The law (nominally) tries to figure out intent behind the data. It is possible to encode any data so it looks like a random stream of numbers; the law doesn't care what you did to the data, just if you intend to use that data to do something illegal. The user's intent colors the data.

https://ansuz.sooke.bc.ca/entry/23 Colour of your bits.

Books are just arrangements of ink molecules on paper. You can't have a copyright on ink molecules, right? They're just a natural phenomenon.

Obviously that's silly — the thing with copyright is not the literal arrangement of ink molecules, it's the work represented by the arrangement.

Similarly, bits are math, but a specific arrangement of bits can represent a work with copyright.

My thoughts exactly. Of course anything digitizable can be thought of as "just numbers"; the point of copyright law is that the act of arranging such numbers is a creative work.

"12 beats" makes no sense, they're talking about melodic phrases of 12 notes or less. You can also have a same sequence of notes, but different rhythms or starting on different beats and it'll sound totally different. And then there's the non-diatonic, or borrowed chords - doesn't even take into account Lydian or Mixolydian mode. I dunno, first pop melody that comes to mind that is an example, Michael Buble's "la la la" line in "Everything" is not in this collection.

Also, for actual complete melodies, I'm not even sure Happy Birthday would fit their restrictions. It's diatonic and within an octave, but the range is from V to V rather than I to I. So if they're assuming I to I, it wouldn't fit unless they supported mixolydian mode.

If you watch their Ted talk, he discusses them expanding to those combinations as well.

https://www.youtube.com/watch?v=sJtm0MoOgiU

Hackers have this idea that they can hack the law. However, judges are not computer programs blindly following the written instructions of the law. They are expected to apply "judgement" when resolving these cases, and they are likely to not be impressed by this effort. I suspect this will have the same effect as if a bunch of people got together and released a massive public domain dictionary and then claimed that authors could not bring copyright claims anymore because all the words they used were in the public domain dictionary.

You _can_ hack the law, just not in the ways that people unfamiliar with law think they can. For instance, Ruth Bader Ginsburg made huge strides for gender equality by arguing for male plaintiffs in discrimination cases to set precedents that would later be used to argue for women's rights. It's less about sneaking loopholes past a judge and more about using the court's own procedure/reasoning to push a certain judgement. If this case can set some precedent about how melodies like this are not protected because of their simplicity, or due to the fact that they can be trivially enumerated, the precedent could possibly be used to strengthen copyright defense cases involving claims over a simple melody.

They are not hacking the law. They are pointing out its absurdity.

The problem with that analogy is that the complexity of melodies and sentences are far from equivalent. A melody is much closer to the complexity of a word. Copyright trolls have been getting away with suing artist who have created original works due to a small section of melody sounding too similar to another work. To use the same analogy, that is like an author claiming copyright infringement, because you used a word that uses too many of the same letters as their own.

    possible 12 beat melodies (1 octave) = 8^12 = ~6.9e10

    possible 8 letter words = 26^8 = ~2.1e11

    (Assume there are 170,000 known words.)
    possible 12 word sentences = (1.7e5)^12 = ~5.8e62

Edit: Used the correct numbers from the article and updated the word/sentence length to be comparable.

Yet tax evasion methods such as the double Irish are considered perfectly acceptable. Given many such cases like this, it feels as if the judges do not have a just standard of when they are and aren't impressed.

Good lawyers "hack" the law all the time. They find novel uses and interpretations for things that exist. The hack here isn't some absolute negation of all copyright claims, but now that all (intraoctave) melodies definitively exist, people suing that they "created" a melody will need to find an even more clever argument that holds up now that it is known to already exist.

For example, hopefully these guys will publish their work on a music service, otherwise it might be argued that they simply created numbers and not music, which a judge could realistically be expected to differentiate.

> I suspect this will have the same effect as if a bunch of people got together and released a massive public domain dictionary and then claimed that authors could not bring copyright claims anymore because ...

They could claim that as much as I can claim that this sentence I just cited ^ above belongs to me, it wouldn't change the reality of things that you supposedly wrote it.

We are talking about the combined work of a dictionary and not the individual words right? Because let's say Wiktionary was in public domain and not CC BY-SA, then if Elsevier went ahead and said that they had copyrighted this material (because it belongs to no-one in particular) and sued Wikimedia Foundation for publishing Wiktionary pages. Would that lead to plausible "judgement" and whoever got more money for lawyers would win? I'm speculation off course, hence the questions and I'm not actually familiar with the American copyright system.

I.e. You can summarize, you can quote, and you can paraphrase, but if you copy the thing in it's entirety, bind it, and sell it as a production of that work, you've violated copyright. If you copy the entire literary content, but leave out all the other trademarks, illustrations, etc, you'd still be likely to be held in violation of copyright as long as the amount of word for word reproduction of these original text in an unaltered form approached unity.

However, as far as I'm aware, you could technically get away with reproducing another book in it's entirety if you quoted every piece it of it, but there were other substantial chunks of content around said quotes. This hasn't typically been done for practical reasons, (as you'd arguably have two books worth of content in one), but would still technically just slide in as fair use as far as I am aware. Though I'm fairly certain anyone would just encourage someone seriously attempting to do that to just get permission, or publish a standalone work with references to a particular edition of the work in question.

The absurdity with music related copyright infringement is that the copyright holders assert they own a copyright to the underlying melody. The foundation of the composition if you will. It should encompass the work as a whole. Either that or the particular scoring. The example I gave above of a book that reproduces another book as part of it's composition would roughly parallel another song composed with the same underlying melody. A book containing a repro of a book in the sense of set theory is still a different composition than the original work. So too should be music.

Unless I'm flawed or unaware of some extra nuance about these things.

Quoting generally limits you to 10-20% of the entire work quoted, and you must add a lot of detail to justify those quotes. You could maybe get around it by writing 100 books each that quote 1%.

Even then you might get into trouble if it can be shown your quotes are in order and thus you didn't apply any creative effort in choosing it. (out of order with page references so it can be assembled might be fine).

All of the above with a ton of maybe. It all depends on factors that are not entirely settled and subject to change.

>All of the above with a ton of maybe. It all depends on factors that are not entirely settled and subject to change.

That's the part I object to frankly.

For a second I had a hope that there were actually guidelines that helped clarify it, but even with the 10%-20% rule, yet if one managed to write and publish an entire set of pages enumerating every semantically correct page, we'd again be in the same problem.

As the process of creation becomes more available to the everyman, I can see that copyright is going to be a thornier and thornier issue. Especially if smaller and smaller chunks are going to be deemed as an acceptable basis for infringement without taking into account that the commonality of those smaller chunks being allegedly infringed upon must have the quality of being arranged in reference to every other part of the composition as a whole as well to count as an infringement of you're even attempting to achieve a reasonable outcome of copyright. Which means that you couldn't claim infringement merely for the presence of a work in another, which I'm fairly certain copyright holders would find unsatisfying because it would sharply curtail the value proposition represented by going after people for purportedly "infringing" in hopes of garnering a settlement, which is what it seems has become the norm rather than the exception in IP circles nowadays.

You can hack that judgement with money.

> "if the artist being sued for infringement could have possibly had access to the music they're accused of copying—even if it was something they listened to once—they can be accused of "subconsciously" infringing on the original content."

Now, they just need to release an album called "Every Song Ever" and start having that played on the radio so that they can argue that every musician can claim in court that the version they were subconsciously inspired by the Every Song Ever version.

Not before that radio show is over which might take a while.

On a more serious note, if that project was ever taken at face value, that is as a collection of music and not as a piece of mathematical performance art exploring the theoretical boundaries of copyright, they'd get into quite some trouble: somewhere on that drive there's a copy of Strawberry Fields, perhaps even the Happy Birthday song, and we all know what that means.

perhaps even the Happy Birthday song

Not to be pedantic but at least this one was finally put to rest in 2018, in the US that is. And it's EU copyright expired in 2017 [0].

Copyright law is crazy but every once in a while a legal case comes along and injects just a tiny bit of sanity back into an otherwise insane system...

[0] https://en.wikipedia.org/wiki/Happy_Birthday_to_You#Copyrigh...

I'm pretty sure that 2 works independently created that are the same, both own the copyright. An infringer has to have had access to the work he/she is being accused of infringing.

In Katie Perry's case it was claimed because the Youtube video of the accuser was viewed 3M times on Youtube she could not say she never had access.

An algorithm can't have access, so imo both parties would own the copyright (if an algorithm can own a copyright)

Reminds me of this work: "On 12 Sept 08, German Avantgarde musician Johannes Kreidler registered – as a live performance event – a short musical work that contains 70,200 quotations with GEMA [german copyright collecting society] using 70,200 forms."

http://kreidler-net.de/productplacements-e.html

This resonates with feelings I have any time I want to work on a music based software project. It feels like a minefield even contemplating ideas in this field; like anything successful that analyzes music will get you sued.

I don't think this is going to put a dent in cases of existing copyright holders making claims against people coming up with similar sounding works.

Firstly, there's an enormous corpus of music out there that precedes this effort, and those copyright holders can still make claims.

Secondly, and perhaps more importantly, how would someone prove that a musician copied the work from this algorithmically generated collection? These haven't been number one hits or in widely available collections. Someone could not conceivably argue that a musician derived their work from this due to familiarity with the melody.

I wish the copyright problem would go away. But I really don't see this as a solution.

A large number of musicians and software designers need to collaborate on a project that is much more permissive with regard to sampling, covers, arrangements, adaptions, etc, all the while preserving the rights of songwriters who just happen to write a number one hit song to get paid for their work.

The whole thing needs to be lucrative enough that composers, performers and the tech companies involved can afford to flip the middle finger to anyone exhibiting corporate greed.

>how to distinguish a solution like this from a traditional record label.

It could be a traditional record label, just with a rule in its charter that all works are released into the public domain ten years after recording (instead of the regular 70+ years). This should still be profitable to run since you have a decade to profit from any song.

Maybe even immediately releasing into the public domain and just operating from the profits of live performances could be a viable business model for a record label.

Release after a certain number of years is one possible way forward. The problem I see with that though is you want as much excitement as possible about your number one hit in the first few years. Covers, adaptions, arrangements, performances all add to the excitement around the thing you are trying to sell. You should want those.

My personal (not very practical) opinion on it is that any derived work that is musically distinct should be considered a separate work. If it isn't the same tune, it isn't the same tune, even if it is similar. If it isn't the same performance, it isn't the same performance.

In fact, the only thing that should be copyrightable by a composer is a specific arrangement of a melody, not the underlying melody itself. (The problem of course is that this is subjective, and not something the law or technology is well-equipped to decide.)

As for record companies, all they should care about is people ripping off their exact CD's and putting them on YouTube for free and not paying them. In my opinion, it's greed to want a cut of a performance/arrangement/cover of a melody that someone else put all the work into.

Give enough monkeys enough pianos and time and they'll eventually write Mozart's Piano Sonata No 16.

Interestingly, it wouldn't be eligible for copyright[1], but somehow the algorithmically generated work would be?

[1]: https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...

> Give enough monkeys enough pianos and time and they'll eventually write Mozart's Piano Sonata No 16.

Sure, No 16... No 19 is right out though.

An "8-note 12-beat combo" is nowhere close to being a complete catalog of every possible melody. Not even in pop.

Maybe, can you give an example that does not fit into this?

Anything that uses 9 of the 12 tones in a Western standard equally tempered octave, for starters. So, if it changes key at any point it is very likely to find that it uses 9 tones or more.

Anything written in a melodic minor scale will use 10 tones, as the 6th and 7th of the melodic minor scale are sharpened on ascending movement but not sharpened on descending. Assuming it uses every note in the scale of course - not all melodies do. But if you start off saying 8 tones per octave you're probably not making space for sharpened 6ths by ignoring the 4ths.

Just about anything written by Schoenberg in accordance with his theory of tone rows and atonality, where the idea is to use all 12 tones as a fundamental component of your melody so that you're not constrained by the notion he didn't like of a tonal centre. Okay so this isn't a popular genre, because it turns out most people who aren't Schoenberg really rather like tonal music. (Yes, that's an oversimplification of his ideas, he did a lot more for music than that).

I think the author of the article has exaggerated the claims about "every possible melody", because given the people who did it are musicians themselves they must be well aware that they've ignored a large swathe of music, and that's before we've even considered tuning systems which don't have 12 tones per octave.

You're missing anything that uses notes outside of a typical diatonic scale. Example: The first 2 notes of the popular piano piece Fur Elise. It uses an Eb. The piece is in A-minor/C-major and Eb isn't in that scale. That piece also uses a G#. The addition of the Eb and G# means you need at least 9 tones to represent that melody.

Many Spanish songs use what's called the Andalusian Cadence which will use the equivalent of that G# in melodies too as well as the non-sharped note.

Any blues singing will use that "blue" note that is off the diatonic scale.

A song like Jamiroquai's Picture of My Life has a little color note that feature prominantly in the melody in the opening line "... I can follow through." The note on "through".

Blues has more than one accidental: https://www.basicmusictheory.com/img/c-blues-scale-on-treble...

Blues scale is also 7-notes instead of 8.

I play lots of blues guitar, I am familiar. That chart is basing it off the C minor scale which would correspond to the Eb major diatonic scale.

If you play an Am blues scale which is, you'll see how the blue note is really just one note (it's a pentatonic scale with note between the 4th and 5th added): it would be A C D (D#/Eb) E G A. That D# is the blue note.

Though you asked for an example... but maybe its best to show a counter-example instead? Here's the Imperial Death March played in a Major-scale, instead of a minor scale.

https://www.youtube.com/watch?v=B9MShtCg4fk

If everything was in an 8-note (no accidental, transpositions only) measure, they'd sound like "Vader's redemption".

To sound sad, you MUST shift to a minor key. A minor key has sharps and flats to create tension. Any "sad sounding" song cannot be properly played in a major key.

> A minor key has sharps and flats to create tension.

The natural minor or Aeolian mode doesn't use any notes outside the diatonic scale (probably what you meant by "sharps and flats"). It's very possible to write sad music in the natural minor. REM's "Losing my religion" for example.

> To sound sad, you MUST shift to a minor key

With sufficient skill you can write sad music in any key or scale, there's no hard and fast rules here. Tonality is only one of the elements you can use to shape the emotion that's conveyed, and it's all quite culturally relative too.

Most keys contain sharps and flats so they don't necessarily create tension. It's a note's relationship to the tonic that does that, and what precedes or follows it.

And as we all know, sad songs say so much.

"Maria, I just met a girl named Maria" in the first two notes; it's Lydian. Or if you pretend it starts on the IV, then the 7th note won't fit for being outside the octave.

Yoda's theme from star wars... first three notes of the Simpsons theme...

Pretty much any song by Nirvana... like Lithium, major 3rd, flat 6, flat 7 - that's not even a normal mode.

"La la la" from Buble's "Everything" for an easy listening example.

If it is all possibilities of 12 beats and below, then presumably it includes a measure + a partial measure of even 7, 9, 10 and 11-beat measures.

But for an example: "Turn it on again" by Genesis is a 13-beat verse. The chorus I think is maybe in 4. You could turn to King Crimson for longer beats.

Shifting meters (Stravinsky's Rite of Spring), quarter tone music (Charles Ives), other alternate scales (check out Okros Ensemble), Take 5 (5/4 meter). The title is absurd because there is virtually an infinite number of melodies.

Anything with a sixteenth note or faster.

Anything not in 4/4.

Anything not in C major.

Duh, but that was not the question.

The question was how common that is in practice? Take the key recognizable seconds of top # songs of the last # years. What percentage will or will not fall under this algorithm (possibly after normalizing the key)?

Not 100% seems obvious, but is it 90%? 50%? 5%?

I went to Billboard's hot 100. I picked the top 10 songs.

The Box: Roddy Rich - It's 4/4 and the vocals use triplets and 16ths; basically monotone, very few pitches --> does not fit

Life is Good: Featuring Drake - 4/4 with triplet phrasing; basically monotone, very few pitches used --> does not fit

Post Malone Circles - 4/4, 16th note vocal phrasing at points, no swing feel, the main melody and chorus are free of accidentals, it uses no swing feel --> does not fit

Dance Monkey: Tones and I - 4/4 with swing; accidentals in the chorus (chromatic descent) --> does not fit

Don't Start Nau: Dua Lipa - Literally all triplets over 4/4 in verse... --> does not fit

: :

ugh.

I can't listen to any more of these awful (IMHO) songs, but so far 0% fit "4/4, 8th note divisions, no accidentals" that was used to generate the 68B patterns.

The scary part of this (and Adam Neely does a great segment on it) is that the person simply has to have a copy of the work to claim the copyright. It reminds me of genetics companies patenting "random" sequences in a hope that some might contain some value in the future.

What happens if the algorithm creates a melody that is already protected by copyright? Are they prepared to defend themselves in court? Are they willing to take the "offending" piece out of the library?

The algorithm has almost certainly has done so many times. But it just writes numbers (in the form of MIDI sequences) to a file. Numbers aren't copyrightable. I don't think there's any issue of them defending themselves. The bigger question is whether these numbers can be used as a defense when another musician gets sued.

I only see an 8-tone scale being used so nope, not every possible melody.

Are there actually any popular melodies that span more than one octave?

Also, suppose I take a popular melody spanning two octaves and compress it to one (just c' to c). Will these melodies actually sound different enough to be considered independently copyrightable? My guess would be no.

> Are there actually any popular melodies that span more than one octave?

Star Spangled Banner's melody spans 3 octaves IIRC.

But more importantly, any melody with an accidental would require the 13-notes that include the 5 "sharps" or "flats" per octave.

Every country song is pretty much played in the 5th, meaning the 4th note is USUALLY played sharp. So your standard country tunes are pretty much guaranteed not to be in this set.

Jazz and Blues music has a LOT of accidentals, and they also are syncopated up the wazzoo. Even IF Jazz and Blues music were all in C-Major, I doubt they tried out their strange rhythms.

At best, 8-note, single-octave, 12-note rhythms, cover a very, very small subset of simplified classical music (Simplified Twinkle Twinkle Little Star, and not the original Mozart version either: https://youtu.be/KKCsujeeu8o?t=50) and maybe some simple folk songs like "Mary had a little lamb". But for goodness sake, even "Happy Birthday to you" spans more than one octave...

----------

I just counted it. "Simplified Twinkle Twinkle Little Star" is 14-notes. Even this beginner song is too long to be covered by the 12-notes auto-generated by this algorithm.

So we're talking "Hot Cross Buns" level rhythms... I guess? But "Hot Cross Buns" is 17 notes long.

"Row Row Row your boat" is syncopated. I doubt its rhythm is covered. Its also 26 notes long.

Here's a better question: what song is actually covered by 8-note, single octave, 12-notes combinatorics? I'm having difficulty figuring one out. I'm picking the simplest songs I know of and they still don't seem to be covered.

Many songs have A and B parts. If we allow combining many different "melodies" to create the larger whole that covers some music in that now you only need to find 8 note, single octave combinations. I don't see why we should allow that, but if you allow it.

8 notes in a single octave pretty much limits you to melodies of someone who has been studying music only a few months. Even my penny whistles are capable of more than that after a couple months of study. (a few days if you already know music or are interested in one specific song)

The whole song or even a majority doesn't have to be covered. The Katy Perry Dark Horse lawsuit was over 8 notes repeated twice. Suits have been lost over seconds long samples.

Yes, there are.

And yes, it could sound different enough. If I invert the octaves of every alternate note it's going to be quite hard to pick out the melody as being the same, because I'd have disrupted all the scales and arpeggios that are the parts we actually recognise as forming the shape of it.

The other point the comment you're replying to is trying to make is that Western music generally has 12 tones per octave (even if they don't use them all, a lot of music uses 9 or 10 of them), and other musical systems used in various parts of the world in antiquity and in the present day may have significantly more (or less).

Anybody who limits it to 8-tone octaves and only a one-octave span and says it's every possible melody is just lying or musically ignorant. Which is surprising given who the people who did this claim to be - i.e. they're musicians.

It's likely that the article's author massively exaggerated on the "every possible melody" thing, because there are way more than you can get from 8 tones in 1 octave over 12 beats.

GP is talking about things like the chromatic scale, pentatonic scale, whole-tone scale, etc

Writing melodies with these would comprise of 12, 5, and 6 different tones per octave, respectively. Not 8.

Plus, some minor keys comprise of different sets of tones depending whether the previous tone was higher and lower, so it might not even be the same set of 8 tones within an "8-tone" scale.

Octaves are just a modulus.

Edit: there's also things like moveable 7 note "scales" and other eastern hemisphere things that don't fit western music theory as easily but I don't know them well enough.

Sort of a nit, but if it's only 8 tones, it might not include the minor iv chord. You can hear it in "Creep" in the "Your skin makes me cry" line on "cry." It's pretty common in songs.

you can just transpose it to every possible key

That doesn't work for "accidentals".

"My Heart will go on" (Song from the Titanic movie) is a good example where no amount of transposition will work. The tension of the song comes from playing the "wrong" note. (My __heart__ will, emphasis on "Heart" is extremely tense because its an accidental: not actually in the key that its being played in).

accidentals have special rules when transposing to keep their character so I don't understand your point

This is cool. It's kind of like the project to algorithmically create prior art of every possible idea in order to weaken patent trolls [1].

Too bad they went with CC0 instead of CC-BY-SA – we could have ended restrictive copyright on music and made all music free!

[1] http://allpriorart.com/

There is far more work involved with creating music than simply creating an unique melody. And yet, maybe, pop music can and should be kept simple — like these generated melodies — to have a broad appeal. It's a shame that artists can be sued for likeness if their work feels original and introduces a new tone and feeling when heard.

The role of the artist is to make choices. Every possible thing is the exact opposite of making a choice, and of a creative act, therefore it makes zero sense that anything could be said to be copyrightable about this.

This doesn’t make a lot of sense. Songs are copyrighted two ways, performance - the arrangement, composition and lyrical content of the song, and the master - the actual mastered resulting waveform.

Samples are copyright infringement on the master and using for instance the same chorus as someone else in your song falls under performance.

I don’t get how this helps either.

You're completely omitting written music, which is extremely important in copyright law. Music can be copyrighted without ever being performed—writing it down is enough.

Cool! Who's going to make a browser to listen to all these first?

Just ridiculous. Choice of instrument/rig and articulation alone are enough. If I annihilate "Twinkle Twinkle Little Star" on guitar like Hendrix did to the Star Spangled Banner, I did that and I deserve to be paid if you enjoyed my playing.

You don’t have to do a lot to get a TedX talk, I guess.

Reply on news.ycombinator.com