I would argue that the original author retained the copyright and mine is a derivative work.
If I took a film, and re-enacted the scenes, does my film infringe upon the original film?
This is a bit ambiguous, if the film were presented as parody for example, but you'd still out of courtesy, contact the copyright owners and see if they / their lawyers think it's infringing before spending years doing so
If I took someone's game, recreated it from scratch, and made a youtube video tutorial, "how to make this game including indistinguishable graphics, textures and gameplay" would the resulting video be infringing?
Even more ambiguous than the last example - if the language stated "similar to" and the resultant product was distinguishable in some way, I think it wouldn't infringe.
My gut feel is that guitar tabs is simplified sheet music - and sheet music is copyrightable, because even if the the play style of the artist is slightly different from the original, the arrangement of the music is what is copyrightable, and the original authors deserve at least some of the proceeds from the use of that copyright
The original author. Translation is not creating a new work.
https://copyright.uslegal.com/enumerated-categories-of-copyr...
In some cases original copyright expired but translator copyright still applies.
But if I make translation, then I may be unable to distribute i. But in exactly the same way original author would need to have my permission to distribute translation.
Key on this one is indistinguishable but there are many Minecraft clone type games by fans and Satisfactory is basically Factorio in 3D and the authors admit being inspired by Factorio. The other reality is that its also at the copyright owners discretion to make it an issue. Course if its fully original graphics it makes it harder to argue for copyright infringement I am sure.
I would go to OpenRCT2, which specifically has you copy over the graphics from a genuine copy of RollerCoaster Tycoon 2, and OpenTTD, which includes new graphics in place of the ones from Transport Tycoon Deluxe. However, no one actually knows if this is enough of a difference to matter. It's more about convincing the original creator to ignore it than actually being legally robust.
And then for non-game software we have Java in Android.
There’s a good movie about this premise:
You'd be infringing on the copyright of the script.
In the film analogy, it's closer to Shrek Retold[1](ignoring parody).
[1] Shrek Retold https://www.youtube.com/watch?v=pM70TROZQsI
Obviously there is a difference between copyright and infringement. Or DMCA should take down playgrounds.
Can one make a business by printing copyrighted music score/tabs? No.
Can one educate playing music with copyrighted tabs? Yes, fair use allows with limitations [1], somewhat similar to libraries not ready for digital age.
[1] https://nafme.org/my-classroom/copyright/copyright-law-what-...
It is better to think about whose rights would be infringed if your edited version were published. If you were to publish your edited version, then you would infringe the original author's rights in the original. However, if the original author were to publish your edited version, then he would infringe your rights in your edits.
> if the language stated "similar to" and the resultant product was distinguishable in some way, I think it wouldn't infringe.
The question of infringement isn't whether the alleged copy is distinguishable from the original, it is whether the copy contains elements that are "substantially similar" to copyrighted elements of the original. (Note: "Substantial similarity" is a specific legal term of art in copyright law.)
Unfortunately transcribing music to a tab or a sheet is subject to copyright. Even transcribing the lyrics of a song is. In fact music sheets are a very lucrative business.
Some sheets are also impossible to find because copyright holders refuse to re-publish them and will sue you if you try to transcribe that music yourself and publish it for free.
Copyright doesn’t seem to have much to say about infringing works created in private for private use, that never interact with the market. They’re infringing in theory, but in practice no copyright-holder has ever(?) gone after such infringement, because it’s just impractical to find out about it.
Now imagine an ML model that, when fed audio, emits perfect tabulature corresponding to said audio—without actually being overfit enough to contain an embedding of any real tabs; and where all the training data to the model was public-domain.
If an end-user of such a model applied it to a song, and what the model ended up synthesizing out was a copyrighted tab, then the end-user—and not the creator of the model—would be the one responsible for infringing upon the work, right? But, since that particular tab doesn’t exist anywhere except on their computer, then—if they just used it for practice, rather than publishing it—it’d be a private creation for private use, “invisible” to the copyright system.
It seems to me that, as long as the ML model didn’t actively advertise its ability to infringe on the copyrights of these particular tabs; or guide users into doing so; then it’d be safe to sell the model (or software that embeds it), right? Even safe to sell it as fit-for-purpose for being an aid in learning/practicing music.
In theory all of this is illegal - because you don't buy the music with a CD, you buy a license to listen to the music from that particular run of the CD.
Automated transcription - already a solved problem for solo piano and guitar, and partially solved for more complex source separation applications - would follow the same rules.
So - technically an infringement, but you wouldn't be sued for private use.
Sooner or later this will turn into an actual court case. Until then, the law on automated transcription is ambiguous. IANAL but I would guess it depend on whether a transcription product is sold as a "music ripper" for other people's work, or as a study aid for your own playing.
And of course CD/stream ripper software has never been illegal. There have been various short-lived DRM efforts, but none of them seem to stick for long.
Now it's fairly easy to rip Spotify playback, but no one seems to mind. But if you started running your own Spotify competitor they might - even on a small scale.
You don't buy the right to listen to a CD, you buy a copy of the CD and the music for use as you wish.
Many European countries have an explicit right to private copy, you can copy/transform/archive anything you have for personal use. There is a tax on storage media that goes to the labels. https://en.wikipedia.org/wiki/Private_copying_levy
I believe the US has something similar (you can do whatever you want with the copy that you own) but an American would be better place to comment on that.
What you generally don't have (across jurisdictions) is the right to make copies and distribute them.
CD rippers have never been illegal. They are illegal in a few places if they circumvent copy protection mechanisms and even there it depends on the details.
You can have a look at the legal FAQ of VLC media player that debunks some of the myths. https://www.videolan.org/legal.html
There's a piece of software called 'Capo', for example, that assists with the process of transcription, albeit not generating tab directly, but which analyzes a piece of music and helps you pick out and transcribe notes as tab - and for personal use, of course there's no problem doing so. It should be okay if it went further and attempted to autogenerate tabs. It does chords already, for example.
And writing the tab out, for your own purposes, from a recording you have bought, is not infringing copyright on the original any more than listening to a copyrighted recording you have bought through speakers in your own home is infringing - I mean, it's possible that it could be considered 'format shifting' a work, like copying a vinyl onto a tape, but really it's more akin to just taking really detailed notes as you listen...
And having a software tool that helps you do it seems like a reasonable thing to use, so how could it be infringing?
In a way, if some ML software existed that could generate accurate tab from an audio file, then that software, plus a guitar, is just a really low-fidelity digital-to-analog converter pipeline for turning the original recording into audio. Just the same as an AAC codec + pair of headphones. The codec turns the original compressed data into a series of numbers representing audio samples, which a speaker turns into sounds; the Tab-transcribing ML turns the original data into a series of numbers representing which strings and frets to play, which a guitar player turns into sounds.
So on that basis, it seems hard to argue that the existence of such an ML model is a copyright circumvention device, any more than you could argue that a speaker is a copyright infringement device.
But...
What if the way you trained the ML model in the first place, for example, was by feeding it copyrighted guitar tabs of copyrighted recordings? In particular, what if it was only for a very limited repertoire - say you just trained it on Hendrix recordings, and official licensed Hendrix tabs?
Then, if you hand it something other than a Jimi Hendrix guitar performance, it might not be very good (although I bet the results would be interesting!) - but if you hand it any original Hendrix recording, the same recording of Little Wing that was used in training it, for example - then there's a really good chance it would just reproduce the official tab of Little Wing.
The legal question might be, does the ML model embody the knowledge of how to produce a good tab from a recording, or just contain the memory of, and ability to reproduce, the copyrighted work?
And when it comes to something like music, that gets very close to asking the question of, when it comes to a musical performer producing a song that sounds similar to a copyrighted work they were influenced by:
Do they embody the knowledge of how to write a song that sounds like that or are they just reproducing their memory of the previous copyrighted work?
Which gets to the 'originality' test in copyright law. And applying that in the case of an AI piece of software sounds... challenging.
But on the other hand, it's pretty clear, for example, that when you hand an audio file to the AAC codec and it produces audio from it, that you can't argue that the codec is just reproducing the memory of the right sounds to make, prompted by the audio file - but on the other hand, it's not like the codec is bringing nothing to the table in terms of figuring out how to turn the 1s and 0s on disk into sounds.
But even if you used copyrighted inputs, that raises a tangential question that's even more interesting to me: if you get a human musician to transcribe your song into tabs... and that human musician learned tablature using copyrighted tabs... then is the musician's mind (or anyone who employs it), in some sense, infringing?
Food for thought :)
No band living off performance fees from local bars and restaurants is going to have the time or money to track down and pay copyright owners for the sheet music or performance fees for the songs they’re covering. It’s the most they can do to get a mechanical license for covers on albums they record—and that often would go to the producer or publisher anyway.
Although this kind of arranging can be done quite formulaically, when done well it is usually a genuinely creative process involving plenty of choices on the part of the arranger. I dabble (badly) with an acoustic guitar, and quality solo arrangements that aren't too hard to play are a very valuable thing. The best arrangement I've created for myself could be generously described as mediocre, despite being based on the official sheet music.
The copyright lies in the musical composition itself. A graphical representation of that musical composition, whether in the form of classical sheet music or guitar tablature, may infringe that copyright if it is not licensed or made by the composer.
Note: This may be more complicated when the tablature is made from a recording by a musician other than the original composer because the musician may have added embelishments, accents, or stylistic elements that are not part of the original composition but are separately copyrightable. In that case then a tablature may infringe on two persons' copyrights, those of the original composer and those of the recording musician.
> So how would you be able to tell whether it was infringing or not?
You tell the same way that you can tell whether any item is infringing: If it was not made by or under license from the copyright owner, then it may be infringing.
The copyright holders think that what Graeth was doing was like taking a audiobook, listening to it and transcribing the whole book and then profiting from it.
It's not copying the performance, sure, but it is taking a roundabout way to copy another copyrighted work.
So it doesn't matter if you transcribe the music by ear. You're still transcribing *someone else's composition" - and they still own the copyright.
This is true even if you get it substantially wrong - but then it's useless anyway, in a teaching context where the whole point is for it not to be wrong.