Machine Learning

1 readers

1 users here now

Community Rules:

Be nice. No offensive behavior, insults or attacks: we encourage a diverse community in which members feel safe and have a voice.
Make your post clear and comprehensive: posts that lack insight or effort will be removed. (ex: questions which are easily googled)
Beginner or career related questions go elsewhere. This community is focused in discussion of research and new projects that advance the state-of-the-art.
Limit self-promotion. Comments and posts should be first and foremost about topics of interest to ML observers and practitioners. Limited self-promotion is tolerated, but the sub is not here as merely a source for free advertisement. Such posts will be removed at the discretion of the mods.

founded 1 year ago

MODERATORS

communick@academy.garden

[D] Is it legal to train on audio that is copyright infringing but not traditionally copyrighted? (alien.top)

submitted 11 months ago by SuperwhizAJ@alien.top to c/machinelearning@academy.garden

10 comments fedilink hide all child comments

I want to train a model for a musician based on piano covers they have recorded of popular songs and therefore own, but because of copyright infringement they still pay royalties to the writers of these songs. Curious if this is legal or not -- I would presume not since the recordings are not being monetized in this scenario but rather fed into a model, and the recordings are still owned 100% by the musician licensing them to me.

you are viewing a single comment's thread
view the rest of the comments

[–] f10101@alien.top 1 points 11 months ago (1 children)

Yes, the recordings are 100% owned by the musician.

The songs themselves though, are not. If your focus is to generate new songs, then while there's a good chance you're in the clear, it is still an unanswered question. You're essentially taking a bet that it's legal.

If your focus is to be able to do a style transfer, i.e. turning music into something that sounds like it was performed by your pianist, then use of the data becomes something that is more easily covered under existing fair use.

I would suggest it may be something of a moot point though: can you actually build a useful musical model with this little data?

[–] SuperwhizAJ@alien.top 1 points 11 months ago

great points. the focus is so he can create more covers with the model. so the latter.

would the fair use norm be that its transforming the original songs/being trained on transformations so it would be all good? the original songs would be used during training as input to the model, but any generations would still have royalties be returned to the songwriters should the musician intend to monetize them.

re your last point: this musician has a lot of data. 30+ hours :P