Suno AI Caught Infringing Copyright Amidst Major Lawsuit | Udio Responds
Вставка
- Опубліковано 8 лип 2024
- On this segment of the Top Music Attorney Podcast, Entertainment Attorney, Miss Krystle reacts to Sync My Music's video "Is Suno Sampling Copyrighted Producer Tags?" The major labels Sony, Warner, and Universal are suing the AI music platforms Suno and Udio for copyright infringement. These platforms are accused of using copyrighted music to train their AI models, which then generate music that imitates existing works. Miss Krystle provides a breakdown of the lawsuit, discussing the implications of AI training on copyrighted material and the concept of derivative works. Sync My Music highlights specific examples of AI-generated music containing producer tags, raising questions about the legality and ethicality of AI-generated content. The episode delves into the ongoing legal battle and its potential impact on the music industry, with a focus on copyright laws and the future of AI in music creation. Join us as we expose Suno and the Shocking Copyright Violations Revealed Amidst Major Lawsuit.
00:00 Introduction
00:23 Major Labels Suing Suno & Udio
00:45 AI Music Generative Platforms Imitating Copyrighted Music
01:40 Is Suno Sampling Copyrighted Producer Tags?
02:37 AI Training Off Of Copyrighted Music
03:35 Suno Continuing To Grow
04:07 AI Generative Platforms Disclosing Sources of Data
05:03 Udio Statement
07:16 How Udio's AI Technology Works
09:13 Suno's Producer Tag Outputs
10:32 Cash Money AP Producer Tag Imitation
11:29 Copyrighted Songs Improperly Used By AI Platforms
12:26 More Copyright Protected Producer Tags
13:21 Derivative Works
14:15 Direct Sampling
14:51 More Imitated Producer Tags
16:08 Recreating Daytrip Producer Tag
17:41 How Is This Technically Possible?
19:36 Did Suno Train Off Of Copyright Protected Audio?
19:48 Modern Day Sampling Machine
21:12 Suno vs Stable Audio
21:54 Suno & Udio Using Copyrighted Content
22:33 Value of Copyright Music
23:55 Should Labels Get Paid By AI Generative Music Platforms
Legal Disclaimer: The information, ideas, and suggestions in this video are not intended to be legal advice. Before following any suggestions contained in this video, you should consult your personal attorney. The speaker shall not be liable or responsible for any loss or damage allegedly arising as a consequence of your use or application of any information or suggestion in this video. Pursuant to the fair use doctrine under the 1978 Copyright Act, a copyrighted work owned by another may be used for criticism, commentary, news reporting, and educational purposes. The use of the livestream contained in this video falls within the fair use doctrine.
------------------------------------------------------------------------------------
🟡Ready to become your own record label? Get started now!:
www.topmusicattorney.com/beco...
💻 #1 Resource For Artists And Producers: www.topmusicattorney.com
⚖ Hire An Attorney: www.delgadoentertainmentlaw.com
📺 Get Your Music Into Tv/Film/Games: www.topmusicattorney.com/sync
📰 Get My Newsletter: www.topmusicattorney.com/news...
📄 8 Essential Music Contracts Bundle: www.topmusicattorney.com/cont...
🎧 Listen To The TMA Podcast On All Streaming: www.topmusicattorney.com/podcast
🟢 Free Stuff: www.topmusicattorney.com/free...
Miss Krystle Artist Links: www.misskrystle.com/links
------------------------------------------------------------------------------------
TOP MUSIC ATTORNEY SOCIALS
Facebook: / topmusicattorney
Instagram: / topmusicattorney
Tik Tok: www.tiktok.com/@topmusicattor...
Linkedin: / top-music-attorney-164...
Website: topmusicattorney.com/
MISS KRYSTLE SOCIALS
Facebook: / misskrystlelive
Instagram: / misskrystlelive
Tik Tok: / misskrystlelive
Linkedin: / misskrystle
Twitter: / misskrystlelive
Website: www.misskrystle.com/
Spotify: open.spotify.com/artist/1g6jA...
THAT ORKO (SHOW PRODUCER AND MUSIC PRODUCER) SOCIALS
Instagram: / thatorko
VIDEO SHOT AND EDITED BY:
Dukes Up Records
www.DukesUpRecords.com
THEME / SHOW MUSIC BY:
That Orko
#SunoExposed #SyncMyMusic #UdioExposed - Розваги
Follow Jesse @SyncMyMusic : www.youtube.com/@SyncMyMusic
🟡Become Your Own Record Label:
www.topmusicattorney.com/becomeyourownrecordlabel
📄 Free Split Sheet Template: www.topmusicattorney.com/splitsheet
📄 Music Contracts Templates: www.topmusicattorney.com/music-contracts
📺 Get Your Music Into Tv/Film/Games: www.topmusicattorney.com/sync
💻 #1 Resource For Artists And Producers: www.topmusicattorney.com
⚖ Hire An Attorney: www.delgadoentertainmentlaw.com
📰 Get My Newsletter: www.topmusicattorney.com/newsletter/
🎧 Listen To The TMA Podcast On All Streaming: www.topmusicattorney.com/podcast
🟢 Free Stuff: www.topmusicattorney.com/free-stuff
Your connections who are "developing" AI are and will be the cause and the demise of all creative human beings. FUCK AI! Just look at what's happening already.
@@TopMusicAttorney wow you are dope!
I feel like the record labels want to own the technology
They do, but it's not going to be possible to purchase. These companies are anithetical to eachother, tech giants will never concede.
@@YoungBlaze You are correct.
What are they going to do with it though? Without a real artist making music, all it does is imitate people and NOT AS WELL as a real musician. They also can't copyright AI music. I think this is more just simply they are being infringed on and defending their rights. I don't know what 'owning' these AI models would do for them, they can just make one of their own if they're going to do that.
AI isn't going anywhere as a technology, and by that I mean it's not going to improve in some moore's-law type way, it's incapable of doing so.
How does A.I. companies stealing from artists equate to labels wanting to own technology? Did you even watch the video? 🤣
who cares. let them own it.. they can recreate all the garbage music they want. i still won't listen to it. i don't listen to it now .. it's trash. today's pop music is made by the same 10 to 15 producers.. . let them retire and be replaced.. the rest of us actual musicians and composers literally do not care.. having these websites shut down helps protect us. so yea we are on the side of the major labels this time.. even if they want this tech for themselves.. i do not care if they replace their 15 same producers who produce trash like Sam Smith and Taylor Swift... lololl why would i care.. i listen to sht like Snarky Puppy , Knower, Keygen Church. I don't listen to major label sht sell out garbage... sorry if i sound harsh.. i'm being honest lol
You make me more and more convinced that the music industry as it is today has to go away. This whole greedy capitalist scheme needs to just be over. Lawyers and greed suck the joy from life.
gee for that to happen dont u think hierarchical economic systems like capitalism has to also 'go away'?
@@CatrinaDaimonLee yes
@@CatrinaDaimonLee Capitalism does not need to go away. MONOPOLISTIC Capitalism needs to go away. Strong anti-trust regulation is needed. An environment where mergers need to go to court to demonstrate how the merger is actually _good_ for the customer (e.g. two airlines with no overlap in routes), rather than demonstrating how it's not BAD for the customer is the first step.
@CatrinaDaimonLee that would be potentially Wonderful. Capitalism means you need capital to partake in the -ism. I.e. if you aren't rich -> you're a slave. Be in denial all you want, but that's the way it is.
@@Recuper8 based
A.I. doesn't need to store data to recreate data. Once it has "examined" something, it is loaded into the dataset permanently.
Much in the way humans listen to a track. You don't need to hear the song once you know it.
The source information is never saved...
This will never be outlawed !!
Please be wary of the real purpose behind the copyright groups’ lawsuits, which may be to gain control over the most advanced AI music models. Undoubtedly, UDIO and SUNO must take responsibility for their misuse of training data. However, imagine a future where AI music platforms are controlled by copyright groups. While people can still create music on these platforms and obtain copyright certification and revenue, these groups would also gain the rights to verify the copyrights of the works produced. As more and more works are created, the real opportunities for innovation diminish. One day, we might find ourselves unable to create any music phrases that pass copyright verification, which could lead to the death of music itself. Given the limitless combination capabilities of artificial intelligence, this dystopian future is genuinely possible.
Screw the labels i will keep on dubbing to Suno. I feel the court will side against the crooked labels. They just want in like streaming.
This is exactly the case… they always sue then come in the backdoor with something that claims ownership but ultimately screwing the artist… rise, cycle repeat.
There may or may not be a case in the training phase using copyrighted music to learn, if copies are made to teach the model. There probably shouldn't be a case for the output, as the examples are cherry picked deliberate attempts to recreate training data.
I, a mostly amateur musician, have learned many songs, probably hundreds of them. I can play them from memory. So, if someone gave me the lyrics for "All i want for Christmas is you", and asked met to recreate it, i could probably play it back close enough, even without practise.
Now, in both cases, i'd say the burden is on the one asking to redo an existing piece, and if such piece is put out on sale, the publisher of the work. Yes, it should be illegal to publish an AI version of "All i want for Christmas is you", just like it is illegal to publish my own crappy rendition of "All i want for Christmas is you". What should NOT be illegal though, in my opinion, is having learned the knowledge how to create this kind of works.
I also write original songs. It would be dishonest to claim that all the music i have exposed myself to has no impact on the output i create. Of course everything i have heard has some effect on what i make, stating otherwise would be ridiculous. Regardless of this, they are my originals. Similarly, when Udio or Suno make a new song, if the song itself is novel (like it is in vast majority of cases), it indeed is a new original, not an "imitation" of earlier works - unless every song made by every artist is similarly considered "imitation" of work that came before them.
As far as the producer tag thing goes, i'm pretty sure that if you asked a human trap producer to recreate producer tags of others, they could get pretty close, without actually sampling the originals. And AFAIK this is what goes on in the models too - they're not actual samples, but rather "memories" learned from the training data.
The real crime here is the awful music released these days. 😂
We're trying to get better.
@@JamesSiggins Trap crap!
In pre AI era copyright infringement cases, does the judge ever ask if the defendant memorized the song or had a recording of it? I doubt if the defendant claimed to have no recording of the infringed song and was only familiar with the song from hearing it on the radio or streaming services that that would clear them of copyright infringement.
A case for infringement would be simple however that isn't the same as sueing a company for studying music and imitating it. Copyright infringement is rather specific. Its not so loose to say no one else can make this genre or use this idea, it is very concrete in that it protects the works themselves not some overarching blanket on preventing general artistic expression or to experiment with sound. Just being similar isn't enough for copyright infringement it has to very much be central to that songs spirit but even then works can directly copy other songs as long as they transform them or present them in a different way so as to transform the work. This is the difference between derivative works and transformative works. Derivative work vs transformative work
In the context of copyright law, a derivative work and a transformative work are two distinct concepts that are often confused with one another. Here’s a breakdown of each:
Derivative Work: A derivative work is a new work that is based on one or more pre-existing works. This can include translations, adaptations, condensations, or other modifications that transform the original work into a new form. Derivative works are created by building upon the original work, often with the intention of creating a new and distinct work.
Examples of derivative works include:
A movie adaptation of a novel
A musical arrangement of a song
A translation of a book from one language to another
Transformative Work: A transformative work, on the other hand, is a new work that adds significant value or meaning to the original work by transforming it in some way. This can include adding new expression, meaning, or message to the original work, or using the original work in a way that creates a new and distinct work.
Examples of transformative works include:
A parody of a song that comments on the original work
A critique or analysis of a book that adds new insights or perspectives
A work of art that incorporates elements of another work in a way that creates a new and distinct piece
Key Differences: The key differences between derivative works and transformative works are:
Purpose: Derivative works are created to build upon the original work, while transformative works are created to add new value or meaning to the original work.
Level of transformation: Derivative works often involve minimal transformation of the original work, while transformative works involve significant changes that create a new and distinct work.
Fair use: Transformative works are more likely to be considered fair use under copyright law, as they add value or meaning to the original work and do not simply copy or reproduce it.
In summary, derivative works are new works that build upon existing works, while transformative works are new works that add significant value or meaning to existing works by transforming them in some way. Understanding the difference between these two concepts is essential for creators, copyright owners, and users of copyrighted materials.
AI trains on human content and no lawsuit will stop this lol.
The "Cash Money AP" tag appearing is like the Getty watermark in Midjourney output.
As an unknown artist I find 6 of my artworks have been trained upon… as an unknown little or nothing I can do about that
That cash money ap was also generated by a computer Noone in the world has a voice like that unless someone found a real example of a troll and hired them to say it
Dj Khalid is not the best music 😂
He can’t play an instrument or do anything, he’s an obese Rick Rubin wannabe
trash in. trash out.
I hope there is at least technical people in the jury that know that these models do not make verbatim copies of tracks, it just creates abstracts. If we don't train on the best data then China will, and already has.. We do not want to lag behind the work of China.
If it's really close, then they will probably have to pay some license fee. Maybe customers who want to make something like cover songs will have to pay higher fees.
Anyone else should not be bothered by that. AI training has to be accepted, at least for everything that's a bit older.
Copyright laws do not carry over international lines. So, if music generation gets banned in the States it will flourish overseas. No different how Russia or China copies Hollywood films or how Hollywood adapts foreign shows and films for Western film/television. Like I said before on another one of your videos, these Labels do not care that their artist material is being used to train AI, they ONLY care that they are not being paid to do so. Do not think for one instance the label gives a darn about their artist beyond seeing them as an ATM.
yep. if Udio gets banned in the US, what’s stopping them from moving their headquarters to Japan where their courts have already decided that training on copyrighted material is fully legal?
a chinese company has one of the best video generation AI services in the world. if these US services get shut down it’s just going to be a HUGE business opportunity overseas. even if US citizens have to use a VPN to access it.
Depends. Downloading music and movies is illegal pretty much everywhere. Countries can still choose to uphold copyright, especially in places like China or Russia, but they can do it and they can more effectively brutally stomp it out if they choose to.
@@BramstukerScott-mb8kv
Stopping "piracy" never succeeded. In the future people might only need to download a music model which they can host themselves.
@NoidoDev I never said anything about stopping piracy. I said countries adopted laws that made music piracy illegal based on the ruling in the other countries. If it gets ruled as illegal in one country, other countries might adopt the law, how thoroughly and heavy hand idly they carry it out and punish the pirates varies, but that doesn’t mean a judgment in a trial like this should be ignored is what I was getting at.
@NoidoDev. Yeah I’m not sure how many people are gonna take the time to download and make their own music using AI. It’s already ridiculously easy to stream or download music for yourself, even without paying. Going that extra step of wiring prompts for individual music seems out of the realm of 99% of the population. I’m not even sure people are going to want to take the Romeo write prompts for a playlist or album, when most people at most just search for an artist they know and stream their music.
This is why we were supposed to get rid of the lawyers. They suck
There is no point in trying to even make lawsuits against AI... you don't understand how advanced and how fast this technology is changing...
Those "models" are just couple of GB's large, they can be transferred to any USB/mobile device and can reproduce all the music in the world, or art, etc...
It's not possible to stop something that's easier to distribute than a fucking movie... And the models will become more advanced, and anyone will be able to run them locally soon without access to the internet.
There is no theoretical way to stop so advanced technology :/
The most sensible legal approach would be to hold people who take material from udio or suno Individually accountable for releasingle pieces of music. This would have to be done in a case-by-case basis with a clear analysis of the piece that demonstrates copyright infringement.
Suing Sudo and Udio is like suing a laptop with music production software that includes samples.
But that would mean flooding the court system beyond its capacity, trying to extract money from people who may not even have enough to make it worthwhile? I think they are after a much more accessible (and ongoing) revenue stream?
I disagree that it would flood the legal system. Case precedent would be based on evidence of copyright infringement based on substantive similarity within the context of present day law.
Two pieces of music may sound very similar which is the case today but no one is being sued as genre, style, chord progressions and voice cannot be copyrighted.
Melody is another beast altogether.
What the labels, in my opinion are attempting to do is to subsume all presently uncopyrighted parts of music under a single category. In this case machine trained. This reminds me very much of the late 1970s when home recording of music via cassette was considered a threat until people in the industry understood the benefits and opportunities.
@@kokopelli314 just to be clear, I do agree with the fundamental point that you're making in that individual users of AI will need to be very careful about what they release (as much as anyone else needs to be already, really?) I'm just anticipating that even if further precedence is established, proving whether any AI generated melody infringes copyright is going to be a resource-intensive process that the labels will want to avoid. My understanding is that they prefer to collect to their money with the minimum level of fuss and bother needed? If they do manage to subsume all uncopyrighted parts of music under a single category, I wonder what would happen from there? Does that mean that the western world could be left behind as an explosion of creativity using AI tools explodes elsewhere around the globe? Hopefully as you say, people will get over their immediate fears and look for ways to set sensible demarcation lines?
No, as soon as a platform gets big enough, they need to filter out things which are too similar to legacy music.
Then the customer can either pay more, can't create what he wants, or isn't allowed to share the watermarked title he made.
Anything else needs to be legal. Musicians and the industry will have to deal with it.
@@johnlacey155 Yes I believe that people will set sensible demarcation lines unless the court's interfere with that process.
As others have said here, basically machine learning is doing something us humans could do ourselves, if we only had more 'processing power'. They're really doing what humans have already been doing 'forever', but on a more expansive and faster basis. Are machines allowed to listen to and process the source music in any way though? What if the music was sourced over the air from radio stations, where no license agreements are in place? Or sourced from online services where no agreement has been executed? Which raises the question, if no license agreement has been executed by an AI company, where do you go legally from there? Would there be there any onus on them to prove that their machines listened to the material in a license-free manner? Obviously if AI platforms produce material that mimics a real person or a real song closely enough, then existing copyright infringement rules already apply. What other legal principles might be brought out, or even introduced, in this case? Surely this is a bigger event for music and musicians than even the Napster/streaming situation?
ML database is like a hologram. Not the same but can be used to generate alternate versions.
This is why we can't have anything nice
if they try to litigate Sumo and Udio on the basis off uncleared samples than this argument would fail in court because the developers know that it isnt actually sampling, its learning and reinterpreting. If a person plays a riff on a guitar that was inspired by someones style, that isnt sampling. So in my mind lawmakers need to find another way to come at this legally. Seems very complicated giving there is no legal infrastructure or law history to draw example from. Right now its the wild west, its gonna take some time to rein it in. They need to focus on the difference between human learning and machine learning. Show the defense the machine has an unfair advantage, or something like that. I cant talk legalese but you get my drift.
This doesn't sound anything like the artist/producer(s) in question LOL.
This lawsuit is dead in the water, they can't prove copyright infringement of a music style or artist vocal sound LOL.
They may be able to buy off judges and juries, thus wining their case LOL.
You can't copyright a style or sound. If a rapper sound exactly like Snoop, there is nothing Snoop can do about it LOL.
Where were they when musicians complained about drum machines and sequences took jobs from starving musicians?
Turned out that we got better and took the jobs back from sequencers and drum machines LOL.
The sampler was supposed to have killed the recording industry, but it didn't.
They said the same thing about the MP3 and Napster and music downloads.
The Pirate Bay hasn't killed anything either LOL.
The record industry is mad because this the final nail in their coffin of control.
They are just crying "wolf !!" LOL.
A lot of the current copyright laws are archaic and need to be updated to better align with modern sharing (online) and technology.
So, artists are expected to create, share, and never get any compensation? Making a living is archaic?
@georgew2014 with technology there are ways that you can still get paid without penalizing the sharer. UA-cam has this system at times. You play copyright material in your video some owners will allow it and simply make money from the ads that pop up on your video. And yes archaic if copyright office does not explore the nuances of working with an AI system to create
@@georgew2014 Expecting income to always come in the same way as it has in the past of any profession is archaic. Learn to innovate new solutions and evolve with the times.
@@georgew2014 Artists get 6% on Spotify. In capitalism, the productive people are poor, and the unproductive are rich.
But you can't sell you cake and eat it too.
With all due respect, esteemed music attorney, everyone on this earth learns from others who have acquired knowledge on various topics. AI, created by humans, also learns from humans. As its creators, we can learn from our creations. It is not intended to use your computer to generate 10 or more producer tags to find a problem with this new technology. Good music doesn't require producer tags. AI is not the problem; humans are. Instead of using our individual creativity to produce new and exciting original music, some people feel threatened by their own creations and misuse AI to recreate others' songs, which was not its purpose. Just years ago, DJ scratching a vinyl record was considered a crime, and now it's an art form called turntablism. When hip-hop producers sampled copyrighted songs without permission, the issue was eventually addressed. Similarly, these minor imperfections in new technologies will be corrected. The future is now.
I can write a song that infringes on copyrights without Ai. I learned to create music on copyrighted music that I like. I don't see a difference other than Ai does it faster. It's on you if you choose to abuse Ai to make a cover song and release it as original. Where does it stop? Go Ai future.🥰
As a recovering lawyer, and long-time musician, I would never use AI to create music. Firstly, I want to continue to make real music. Real instruments. Real vocals. Recorded live in a studio. Maybe programming a few loops / creating some of my own samples / bit of tracking & MIDI stuff. Old school. I realise that's not how a lot of music is made these days, but I'm not throwing away everything I've learned through many years of music study, practice, rehearsals, live gigs, and recording sessions, just so I can prompt an app to do everything. For me, that's cheating. Others are going to be OK with it, and that's fine, but for me it would feel like I wasn't actually "doing" music. But even if I wanted to use AI - say I had an injury that stopped me playing - I still wouldn't because of the risk of copyright infringement. I haven't heard all the music in the world and I'd have no idea if something that AI produced was a direct copy of something else it had trained on. And where does the liability sit in that instance? I haven't looked, but I suspect the terms of use for the AI music apps have a pretty nasty indemnity clause (even if dubiously enforceable). The other issue is that AI might very well have trained on my music. Because I've retained all my intellectual property rights, it means that no one else is looking out for infringement and taking action would be beyond my resources. So I feel like AI is already ripping me and many other musicians off, as well as building the foundation for putting real musicians out of business. It all sucks, actually. This is one of the very rare occasions where I'm supporting the labels (even if they have ulterior motives).
You sound like a Boomer. For shame
it's the same labels that are slowly and secretly taking over spotify and demonetizing and erasing stuff, they want to gatekeep everything, also what makes you think that the same 20 guys (ghost writers/Producers) who are making songs for the biggest stars are not copying left and right and blending it together? 😂using AI sites is not as different as using sample packs and kits because the output is a whole song or ideas from your input and you wouldn't want to use the entire thing unless you are lazy and have no values or creativity. I haven't used it much but I can say that it does help with writters block and it does help a lot if you were not fortunate enough to have studied music when you were a kid-teenager.
Heck just look at will i am or whatever his name is, he blatantly copied song after song and no one ever said anything or made a scandal, he made millions with the stolen songs and here you come to tell me that the outputs from suno are somehow "dangerous" for the music industry, give me a break, besides it's not like suno is going to copy mediocre songs from some of the people who make videos against it.
I don't know... I haven't seen anyone become a superstar from releasing untouched suno songs, and who's suing universal and those guys for what they have done in the past? no one? that's what I thought.
There only seems to be one example in the video, where the text of the tag wasn't the result of text input by the user. In the others, it seems to me:
1. User inputs copyrighted content (text/lyric), infringing on holder's rights and violating Suno terms of service.
2. User "direct the Services to generate any Output that (i) infringes any intellectual property or other proprietary rights of any party", again violating terms of service.
3. User then suggests that the service (Suno) is responsible for the resulting infringement of copyright?
If a user intentionally does this and the result infringes copyright, is the service liable or the user?
Also producer tags are not part of copyright checkers because that would flag songs. One would use such a program in training to ensure that songs are original.
They are saying that the AI listened to their songs and that is illegal.
It's funny that it's the labels and not the artist out here going after these companies, but they want to say this is in the interest of the artists, is it really?
Those law suits don't really sound like caring about actual creators. It's one business machine trying to desperately hit another business machine.
Imagine asking your friend to sing a song you've heard a while ago at a public area, so no one paid for it. And then filing a law suit for that. AI is essentially the same, but what scared the old busineses is the scalability of the technology. Current legal systems are too obsolete to handle such cases.
And since AI doesn't just samples, it can legit sing any given lyrics, it pretty much mastered this human skill. To make it more clear, imagine suing someone for speaking the same language as you do.
I just got an epiphany: Even though bro is a youtuber, anyone can upload audio now to suno/udio, and right when this happened, lawsuits came after. Anyone can be a bad actor/grey hacker. Especially from a youtuber who made a song on suno seeking to destroy it. So they obviously have a vendetta that can be used against them in court to abusing uploading system but don't show their audience from start to finish malpractice-esque.
Record labels said they want their version of this anyways. It's just a license-deal by knocking down competitors with obvious manipulation since the upload feature came about. Notice the timeline.
I think people are equating AI training with sampling. Just cause they're training on copywritten material does not mean it's sampling copywritten material. I believe research falls under fair use, does it not? Please correct me of I'm wrong, but if the AI produces a unique sound, not derived from any other file, it's not a "derivative ' or "transformative" work, is it? So learning the formula of music in general, or any given genre, isn't an infringement on copywrite, correct?
Now I don't know how Suno and Udio are doing it, maybe they're sampling, or maybe their not, but if one were to have an AI pump out unique multi track midi files for vocals and various instruments connected to their own virtual instrument, including voice synthesizer not unlike synthesizer v. Is it really that hard to believe AI can generate unique audio that doesn't infringe on copywrite, and be near production ready? If the Major DAW's aren't looking into including AI, they should be, cause it would give the artist far more controls on output than they could on a web based platform.
The AI is given static at first and required to remove the static. Not much data is left of the song. It's static over static Then it need to generate original songs and guess what is next in a song.
TLDR: This is a messy situation & I'm not sure our lawmakers are capable of dealing with it in a way that truly protects artists/musicians.
As a side note on this topic, I also had the same following same Qs when Midjourney etc came out....
What if they used works that are in the public domain or works that were never "legally" copyrighted? That surely should be acceptable, but still not necessarily allowable to be copyrighted by the AI user, or would it be?
For example (completely not the same topic but still based on law), our govt can't view our emails UNLESS they go outside of the US to another server - which occurs, dare I say, 99% of the time, so what if THEIRS is actually in another country that doesn't have the same copyright laws? I mean, if our own govt can get away with it, why can't these various "trained" AI's do the same?
Suno now has the ability to take a melody (voice, music, etc) to be uploaded to create something.
***IMO without reading the TOS since this became a thing, I believe they will be "training" on that as well, which might become a "possible legal defense" that very well may be acceptable in a Jury Trial among jurors who are not artists/musicians (heck, the fact is the majority of Americans don't even know/understand their BASIC RIGHTS, let alone copyright law.
And what about the songs people created by including their own lyrics, and prompting the type of voice, music, beat, style, etc. Couldn't the creator make the claim that it is copyright work?
Some can't play guitar, drums, etc but can use DAW etc (c'mon, AutoTune is NOT "created by a human" but is a TOOL) to create works?
How can Suno Udio etc NOT be considered as another TOOL?
---> ANOTHER PROBLEM: If they used STREAMING services to train copyrighted works, would that affect the artists/musician income if they were deemed to be BOTS even though they were streamed/downloaded (perhaps even bought)?
Could this argument be used when DSPs etc ban artists for botted streams???
This actually could be a REAL conspiracy of them (like Spotify, especially with the recent info about one artist having more streams that MJ et al) literally working together to screw artists/musicians over?
I'm literally saving up to get your label course, and I've TRIED a few of these "platforms," but I never got what I wanted so I don't even play with them anymore - never intended to copyright that stuff anyway.
Sorry for the long winded comment. Hopefully someone can answer some of my Qs. Thanks in advance!
The prompts would have to be extremely specific with a lot of on platform editing with massive re-roll attempts by the individual who is trying to get the results they are specifically looking to copy. A final, 1 take attempt, for a passible cover is extremely doubtful in all cases. Lets see the working files from Udio account in question.
FLAWED EXAMPLES (they can't name the tracks that and then be like "omg look what it reproduced"..like that person was clearly TRYING to get it to reproduce those as seen from the titles they made when prompting for those songs)
Don’t get me wrong. I’m all for humanity, but did we as musicians all have to pay licensing fees to train off of pre-existing music?
I’m certainly no legal genius so I’d imagine that I’m missing something here.
Bummer. I love Suno.
I really thought they were past this.
Voices can't be copyrighted. Using voices (not their voice samples) of famous artists when you are not specifically marketing the produced music with their likeness shouldn't be a problem I think. Voices may not be universally unique at all.
I think; When generating producer tags or other seemingly sampled copies what happens is :
These platforms possibly apply voice, instrumental separation first to the training material and those producer tags probably ends up in the instrumental section. And since they are trained with multiple songs with those tags they exhibit the known weakness of these generative models to expose the training material when fed with non curated input.
Laying out all the producer tags and getting them all in the output suggests otherwise though... interesting...
Still, I think how the models got their creativity in their output with too many different instruments / sounds / voices and stuff lies in the fact that they probably work sample by sample (44100 times for a second) when training and when inferring hence they can't do things like choosing a key, separate stems and so on.
Legally, Suno or Udio producing copyright infringing stuff is one thing, someone taking that output and releasing it another thing. One can not be held liable until he/she publicly release the infringing stuff. As humans, we all have the same ability of taking a producer tag or something else copyrighted and use in our production. That doesn't make us the target of a lawsuit until we release the thing and actually cause the infringement.
Can these companies be held liable for training (similar to how people train) off of copyrighted work is a hot debate and I'm certainly have the same stance as Ed Newton-Rex who resigned from his work because the company he worked chose to train their models with copyrighted work.
On the other hand, I use Suno, I have even several stuff published generated by it. The way that it would work out is if they somehow kept a metadata that maps the output to the say max N most influenced input training material and proceeds from that song also distributed to the right holders of those songs. Either way, whole music industry should adapt to what comes next with regard to these AI tools.
I have been messing with Suno and have made songs of my own on the website. It is still new and probably needs some development but the overall good of Suno outweighs the bad. My opinion ofcourse.
I have returned so many "no tracks" or just a few second clips with these platforms so I actually do think there are a variety of filters in play.. IMO the labels should cooperate with these AI platforms to insure Content ID filters are in place the same technologies used to prevent copyright infringement on various music distribution sites including youtube. IMO this standard should be enough to prevent any similar songs to protected content. IMO it is actually going beyond preventing copying -- I was hit with copying a song Flitterific and I totally wasn't didn't have it in memory at all when crafting a jam. One element a C5 to F9 slide was similar but not identical, and both tracks had horrible audio quality that was similar thta is all it took to hit soundclouds content id filter. IMO implementing content ID should be enough but I don't think that the merit is simply preventing it they want to get rid of these leading AI platforms and make some money off them imo. That is how I see it. If they honestly cared they would be negotiating with these platforms to prevent content infringements by insuring that filters used are industry standarized used by the other content providers and that there is a DCMA escalation process. Also no ai contributing works ever hit a content ID Filter on any platform .. just saying. IMO if they want to sue a released published work for infringement that is their perogative but I havn't seen any individual cases or calls for joinder on infrginement or class action with multiple instances to substantiate a lawsuit. It may exist I just havn't seen it in my own experiences with Udio and Suno in fact I think their music and the labels music sort of sucks in general and isn't top quality. Is music similar sure genres of music often have similar feels that is what makes it a genre. The labels need to release some good music and stop getting upset about bad quality of music being imitated. If people are opting for badly reproduced 0 brand content over the labels offering there are bigger problems. BAD CONTENT, BAD SONGS, 0 value brands. They might as well be releasing their music through PH.
Also producer tags can't be part of copyright checkers because that would flag all of the songs made by the producer. One would use such a program in training to ensure that songs are original, so the tags didn't get stripped out.
The AI learns music first by imitating with the goal being as close to the original as possible.
Then it is forced to make original works using content ID software, but there seems to be some holes there.
On Udio I've been able to get spot on recreations of Bjork and Johnny Cash, and pretty close recreations of Tom Waits, David Sylvian, and Leonard Cohen.
When they're big enough, they'll need to filter that, and maybe the legacy industry has to help them. Then they can pay a license fee for that, but most of it does not fall under the same umbrella. People don't go there make very similar songs to the existing ones.
The people who want to use it to make some cover songs can pay more than for a normal creation.
@@NoidoDev I think they're *trying* to filter that. My probability of getting it to spit out Bjork's voice dropped after the lawsuit hit so I think they tweaked their threshold settings.
if you seek yo will find, this is the users fault for seeking to duplicate said work mainly, this is not the same as casually using the system. obviously if u dig hard enough into anything u will find what you are looking for, this is no exception and the example here is it the users error for pushing copywrite, instead of trying to make new custom music
What happens if someone releases a model that you can train on your home computer from audio files that you are in possession of? When I purchase an album do I have the rights to train my own unique model?
how will they know 👀
@cerebrumexcrement true, but I still want to know the legal implications.
I force pop lyrics with my lyrical choices. Ai takes my lyrics and learns what I like. It takes a lot of generations to get it, but yes it also learns from my inputs, and it's getting faster.😘
18:23 okay, time for lady Babylon. Enough of this lawyer talk. It is like a leech on the soul
Normally in AI comment sections, people are very reactionary and uninformed. But I'm delighted to see so many people here educated on how AI works. As someone who has used Suno to create more than a thousand songs, I have never encountered a case of a producer tag or watermark slipping into the song. I suspect the people who are getting such results are deliberately prompting it. Similar to how a lot of the viral examples of image-generators copying digital artists were actually using image-to-image generation rather than text-to-image.
Is AI good? Probably not in the long term. But here's the thing, AI is very transformative, it isn't actually breaking any laws. It's going to require a huge paradigm shift to fight AI, and even then AI will probably win because it's so fast. And even a moderate victory in one part of the world would be an opportunity for a more short-term thinking country to become the AI-superpower of the world.
How do you knoŵ they actually got busted? I didnt hear no news yet
You clicked on the link, so it worked.
Really interesting to see this subject properly scrutinised. Keep it up. 👍
Producer tags can easily be compared to AI output by analysing with audio editing software.
Also producer tags can't be part of copyright checkers because that would flag all of the songs. One would use such a program in training to ensure that songs are original.
@@abram730 I was referring to editing softwares enabling detailed analysis of the audio wave frequencies on screen. The fact the these song gen companies don't refer to it as 'sampling', does necessarily mean it wasn't so in this instance. Producer tag is an audio equivalent of a company logo, or a signature. If an unknown graphic artist uses the likeness of anothers signature on his own work, owner of that signature has a case.
Producer tag on any work other than producers own, without permission isn't only a copyright issue, it's fraud.
@@mikeomolt4485 I'm talking about content ID software. If that software flagged producer tags then that would be a problem as the same sample is in many songs.
Thank you so much for covering this! Such an important topic in the music industry right now.
13:11 interesting. So the only way you can add a hip-hop tag (and lyrics) is if we upload audio.
Disclaimer: We don't know what music the user uploaded (to generate the music) nor do we know the lyrics the user fed it.
I can feed it a song from a artist and add lyrics that say a producer tag and get the same results. If users fed it copywritten music we are also unsure if that will be added to it as well.
Having a native copy is better because early stages of training almost need a local copy. Your first training is forgery training. The same song covered in static needs to be run over and over until the AI gets it and what it creates when successful, is a forgery. Not a copy but it is attempting to make it the same. Is it illegal to make a forgery if you don't sell it or pass it off as the original?
One could argue that suno used copywritten work in these early stages and producer tags being in many songs are not part of copyright checking because they'd flag all the songs with the tag. That Suno uses these copywrite checkers in their training to ensure a song is original, and thus missed the tags.
Suno will need to argue that the AI only listened to the songs and that no illegal copies were made.
Thank you very much for this (and all other) informative and important info.
As a user of this platforms, and a musician, I would love to pay more for the copyright agrrements, and let the amazing being on earth that created amazing stuff.
The creation owners deserve to have this credit!
And I truely hope that Suno and Udio (and all other companies that are doing the same in every industry) will do the right thing!
Good luck and keep with this great work you do.
Greatings 😊
Recording artists have been imitating each others works since the dawn of recordings and it has rarely been an issue because most copyrighted music is owned by a handfull of companies so even if an artist repackages another song from the catalog and sells it again the copyright holder still makes $$$. The problem is when any newcomer (AI or human) who is not signed to one of the major labels tries it then the major record labels come down on them like a tonne of bricks because they are not getting their cut.
Biggest mistake Suno and Udio made was not signing deals with the major record labels but that could easily be fixed by signing a deal with them.
Therefore the most likely outcome is they will settle and pass the costs on to their own customers.
"Biggest mistake Suno and Udio made was not signing deals with the major record labels"
Not signing a deal with the devil is definitely NOT a mistake.
@@MindTripExperience
Correct. Steamrolling them is the right way. They would just have spied on them and tried to slow them down.
The legacy industry can provide some filters or detection mechanism, like they probably do for UA-cam. Then these platforms can filter that out, either not allowed it to pay for it.
What I'm talking about is the creation and release, not the training.
Gasight on. There is no copyright infringement with the A.I. models. If Suno or Udio even take these lawsuits serious, the whole thing is a pony show and they have already decided from the beginning to team with these labels. They just don't want their subscribers to think they did it willingly. Besides that, A.I. don't need the big label's music to function as it is. The serious users just want backing tracks, they don't need A.I. to give them the whole song. Working with backing tracks from old uncopyrighted music would be even better for the future of music although I think most people who put music on the web don't care if A.I. is trained on their songs as it does now. We are the ones who will benefit most from the A.I.
If does not copy works. It generates no music that someone can recognize as their own
What about the singers who have not consented to having their voices being replicated? You can't just generate a track in Suno with Carrie Underwood randomly performing the lyrics and then make it your own commercial release...
Something interesting is that the voice changes based on how you write. The voice sounds a bit like you even though it has never listened to your voice if you type how you talk.
Suno knows what you sound like even if you never recorded a song, simply from how you word things.
Change how you write and the voice changes. Want a Jamaican voice? Type Jamaican.
It speaks large numbers of language.
I'm actually really impressed with Udio's new measures for checking their generations and lyrics.
The new moderation system is good enough that it's making the site harder to use.
But I'm glad it does what it does.
As a song writer I always copy and adapt the best licks, lines and chord progressions I can. Paul McCartney admits to copying many artists including buddy holly.
@andrewbrock8920 Yeah but your not coming up with 200 progressions that are similar in an afternoon of time.
@@nwilt7114 With enough human artists you are.
So basically Suno AI is doing the same exact thing that the same ten or eleven producers hired by the labels to derivatively and formulaically 'create' for the 'artists' out there. Keep in mind that the labels have constantly been pushing for 'I need artist in this genre that sounds like this, because we know this sells and gives us our return on investment.' The lawsuits seems more a preventive measure against letting the public peer behind the curtain that no real creative innovation is going on anymore within music (or has been in the past few decades).
Thank you for the great videos! Question related to this and some of your other videos: should I register a producer tag with the copyright office? If I do that and then use the tag in a song, would that be a derivative work or sample?
It should be made impossible to enter producer tags, as they have already made with artists names, of cause they have trained them with everything Internet has, yet they have to filter inappropriate results that are mimicking copyrighted stuff. I love Suno and want it will be safe for everyone- people who create music themselves and those who use AI tools also. ✌️🌈
If there are ways for AI to give power to the actual artists or artists owned labels.... that is the tectonic shif that AI brings but with the right human intent.
The way things stand, that's looking like a big IF.
Remember when MP3 came out? The losers are back!!
What sucks is that the labels are probably going to get paid and all the independent artists whose music is in the training data won't get shit.
Which is interesting if they pointed there AI at SoundCloud!
The record labels are definitely greedy, but also I don’t support tech nerds using copyrighted works to train their AIs without properly compensating the artists and song writers who came up with them. I think we 1000% need regulation around that.
If you don't make music then people can't listen to it.
To be honest, you can get those tags without going through the original tracks of the producers. I remember 2016, when every producer was dropping the metro booming and 808 mafia tags and their beats as if they were normal transition effects 😂
These lawsuits are stupid.. WE as humans learn in a similar way AI does.. you build patterns in your brain and "your music" is based on all you have heard.. so if you can sue an AI you should also sue every musician for their material. Secondly, why should I not have copyright on material I have created with AI? I did the job.. I asked the AI for help.. I still do the promts, and actually initiates the creation. And music writers losing jobs? SO? Very few today hires in a live band to a party, instead of using prerecorded material.. Just because you are a creator you are not any more special than a mechanic.
UA-cam has millions unauthorized songs and movies in their entirety how come youtube isnt being sued
Midway through the video and I'm already offended by that response from Udio. So, with their remarks about originality in their AI system, this extrapolates to say they're not copying other works. Fine. Sure. Then, how do they not acknowledge the cases when it has blatantly replicated existing works? They could at least say it's a glitch not a feature. Instead, they take the gamble that people can't draw the connections back to preceding works? That new cash infusion will likely have to be spent on defense attorneys.
It's so easy. Just invert the phase and play the tag from the song and the Ai generated thing. If it starts to disappear it is NOT generated. It's the same sample. And by the way. Even if it doesn't cancel out, they DID train it on it to be able to generate it in the first place. This is ABSOLUTELY illegal and against the copyright law. They don't own this. They are not even allowed to listen to it and learn from it because it's not theirs and it's not cleared. Even if it's from Spotify it's not technically cleared because it's not learning for educational purposes because it's being used for commercial reasons.
So if you buy a CD album and listen to it, then you have violated the law?
Imitation is the best form of flattery. Go Suno.😊
If it's EVER streamed it's NO DIFFERENT than RADIO!!
If you make money then THAT is the DIFFERENTIATOR!
Use your DAW, "STEM" it out and WALLAH you fixed it.
lastly that's also important: As they're making their own version to cut out competition and license your the melodies, it's a smart business move, but artists don't really win since melodies overlap a lot of times since there's not many to go by with an pop-focused E key over a syphoned flow that few like and buy from (music can be vast, but selling music tend to sound similar). This could, deductively, in turn, turn our music into South Korea, where few actually succeed by design, but in our case, by IP syphoning. If you think about it, you don't lose sales if your music is made by you. You can only lose sales if someone is copying the song and profiting under your name. People buy from the artist, not just the art--otherwise the unsung best music makers underground wouldn't be underground. Clout plays a big role in music. The biggest role, since even people with terrible (MGK) or mediocre (jayz) music can be seen as the best or top tier.
by tearing down Suno for shekles, you're basically helping labels tear down you and your ability to be more creative than them for the first time. They have their own AI in the works. So you'll be in a lose-lose scenario as an indie if they choose to "not" make their licensing system public, and literally be like south korean artists, and design their successors.
And since all melodies are public domain years ago, our copyright laws over music patterns died then (it's up to the user to profit off of obvious copyright after generating it--this is not a legal issue a generator should face if they're not profiting off of the content generated, only the generating user can). Check their ted talk. However, producer tags is an issue, but it's short in time and might be thrown out like watermarks were thrown out for artists from stable diffusion. However, record labels are devious like subprime loan sharks (ironic, since they have advances), so they'll use litigous means to get their bottm line, and have done so historically since the early days of the 360 deals with their own artists, practically tour slaves that must pay most of the expenses from their portion with the loan, and the loan last time i check go to the artist directly like a check, which means taxes take much of it, which makes them further into a debt most cannot recover. It's a high level game they're playing.
Suno claims the generator can make money off of what they generate, however, anyone could use it technically. Congress said anything made by AI is uncopyrightable, though could go in the weeds depending how much of the art is actually made by an artist vs. what's generated. Hopefully, they'll be a way to tell in the future (What was an original prompt vs. the artist).
these greedy record labels have done so much harm to artists, producers with their shady business practices, bad contracts and simply just taking publishing that they dont deserve being as though they dont create music, there is no justification like Prince said for any label to take a dime from a songs publishing! they are suing for music they own but the artists who made the songs dont own their own songs? they wont give a dime to the artists who they are now acting like they care so much about INTEGRITY in the music biz lol i hope they LOSE this lawsuit because this could be wat we need to change the old traditional music model, destroy and rebuild WITHOUT the labels going forward! as far as AI music goes if you are truly musically talented then you know that no software can compare to your human creativity or talent and you shouldnt fell threatened by udio and suno! industry rule number 4080... RECORD COMPANY PEOPLE ARE SHADYYY!! As a producer i definitely dont want my producer tag in a ai music database but i think the labels should suffer especially if ther artists wont get any of the money, artists need to start having a backbone and stop letting labels own them!!
Claiming 'new ideas' by copying (or training on) existing sounds is a contradiction by itself.
The term AI is getting overblown to something which it is not.
So how is chatGPT different (which ALL of AI has basically been trained from)? It was trained from same places. Music, writing, poetry, encyclopedias, internet, etc
I don't think it is different. There just no companies going after them. Major labels are generally aggressive.
@@mrr5835The Authors Guild did file a class action lawsuit against OpenAI. Earlier this year a federal judge in California dismissed parts of the copyright lawsuit. They may have filed an amended complaint since then
Difference is that there is no industry to fight against on behalf of writers/whoever it's replacing. Music is literally owned by 3 massive companies, due to the cringe publishing deals going on for the past decades, and useless artists who can't produce their own stuff.
I suspect its only because there is enough money being made off of music. All the people potentially ripped off by other Ai are spread out vs controlled by a few money hungry companies
Zero difference
The producer tag if it is repeated enough in music the AI was trained on, the AI will incorporate it into parts of songs that sound like the sources that incorporated the tag.. Im guessing Suno trained theirs on a small sample of copyrighted material to keep the sound pure, cause it would rule out the tag if it had heard enough content like that.. I believe that its actually harder to trace content to particular artists if you train on a lot of content, versus little.. Udio's music does tend to sound like popular musicians, in fact I made one tonight that was singing like the lead singer to "the knife" and all I ever do is drag out content that sounds like it comes from specific artists, I have a knack for it. But UDIO may actually be better off than Suno as UDIO's model is more refined, and Suno trained theirs mostly on modern music.. UDIO trains their model on all music.. You should look into how diffusion and neural networks work, also how transformers is used to help the model train itself.. The chance of recalling actual is less likely than replicating the singing and composition style of a musician, which could be more damaging, but technically is not copyrightable cause its not storing the works, its only imitating the musicians more perfectly. And if the AI is better at doing a musician than the musician, it doesn't mean the musician is talented, it means they've spent a lot of their career being recognizable and distinct, it will also hold for specific styles of instruments and loops like band in a box produces. This is no fault of UDIOs, its model cannot discern general music from copyrighted, even when the singers and musicians in generic works are singing like commercial musicians or using their instruments.. But the more music you throw at the model, the more generic its content will be, cause it will have become like stones that have become weathered, smooth and mire refined, devoid of any details.. What UDIO may gave done wrong was the set if content they used to train the model may not have been general enough, it may have been trained mostly on recognizable musicians.. One thing you will notice is a lot of their progressive rock sounds like early genesis and yes.. But how many musicians on the whole made progressive rock. If the industry was the only source, how is the AI to know what a indistinct singer in progressive rock is, if all the singers tend to sound like phil collins, peter gabriel or jon anderson.. How much progressive rock contains generic voices without a style? It could be that there just wasn't enough examples from the past..
My guess is that content ID doesn't flag producer tags, and a content ID system was used in training
How do we know someone hadnt uploaded these actual sounds into suno..tou can uoload your own audio into the app
When Udio said that "every new technological advancement in music was initially met with apprehension but ended up being a boon to artists, record labels, and the general public" they forgot that every new technology killed a part of the traditional music market. Human attention is a finite resource. If generative music takes away much of that resource, traditional musicians will have nothing left but to either move into generative music as makers of samples for neural network training, or leave commercial music for smaller niches.
I think a higher premium will return to the live music experience, especially of a more "unplugged" variety. Everything old will become new again.
@@TimGallant As soon as something becomes popular (or as they say nowadays becomes a meme), a neural network will be trained on it and in a week every schoolchild will be able to generate the same sound on their smartphone.
You're right, but no one is entitled to be paid for their passion.
@@neominimalmusic As long as the originator gets his or hers first,,,who cares
glad i found you! starting my own label and about to buy your course. thx.
That's great! Please feel free to reach out directly to me at krystle@topmusicattorney.com.
It's definitely easier now to create commercial sounding music. That must be copying commercial music. And if users ears tend toward the sounds that sound, "commercial", they must be "copying" the labels artists... This is going to take a long time to figure out, in the meantime, labels want to be sure you can't get access to their markets. And of course artists on those labels will be in alignment as will their attorneys. Judges will be faced with an alignment that will be challenging to deny. Doesn't mean it's correct.
The way i understand copyright..if you create a exact copy sure thats infringement..if you boworrow a cord a melody thats infringenent. There are thousands of songs similar in structure tone beat etc..answer this question..if 2-3 seconds of one track is used..and 2-3 seconds of anothet track is used etc. How can possibly be infringenent? How would any human artist use anything if thata the case?. I totally agree the replication of ones voice if thats the case should be stopped..how will you prove that these companies took entire tracks and put them in their software or their models? ..I believe its inpossible..furthermore platforms that's host real music such as youtube have sofisticated detection technology to flag actual works..arent they flagging them? I woiuld also say..courts have also ruled on brief usage. The estate of murray leinster, a noted sci-fi author sued paramount pictures because they clained exclusive usage of the phrase " may the force be with you" from star wars. Paramount won the case..just how much of a track is illegal to recreate? Are the courts gonna provide a definate answer to that? Again how in the world can a human reuse anything of every second of a song or beat or whever is protected?..Also congress has mor passed any new laws updating copyright with this new tech?..I am in total agreement nobody should be able to sell someones work in its entirety. But if the model is taking 2-3 seconds of litterally millions of sounds beats etc. They shouldnt be found liable. Im seeing a example of 2-3 seconds of recreations..that production tag sounds like it was generated by a computer who has such a crazy voice? If that was also generation
It's all pretty clear !
I think it is quite obvious AI has seen you right the actual producers name in the lyrics and it happens to be a song where the producer put their own name in those lyrics. So AI will come up with the same/similar sound for the one line. I am willing to guess that if you wrote 'Daytrip on a summers day', there will be no stutter.
Artist are cheering about this, yet don't realize they had better options with suno than a label on this. They practically can copyright every melody if suno and udio are removed. Even though a researcher has made every melody possible to make in under 2 TB public and free to license.
Love your work. Keep up the great content. We need to learn to remain creative without AI. Human creativity is what keeps us special!
Udio will not let you copyright material...This is why i use Udio.
I’ve been playing around with Suno for about 24 hours and I’m very impressed with the results. As far as it being derivative… I’m finding that it’s showing me how very generic a lot of music already is. Take symphonic power metal for example. The genre was pioneered by a handful of bands like Nightwish, Xandria, Amberian Dawn… and every other band in the last 20 years that has a similar style has been “influenced” by the originals . A lot of it is still great and has plenty of it’s own originality and artistic integrity, but the fact is that to achieve a true banger of a catchy song, you’re going to return to that original formula that was proven to work by a few and repeated over and over again by many.
So what is the user actually going to be using the AI generated music for? Personally, any of the audio I receive from the AI isn’t going to be re-upload anywhere else in any form. I play guitar, keyboard and program my own tracks from my huge collection of virtual instruments and sound libraries. I might take a large section of the music from the AI and use it to recreate my own track, played and programmed by myself. The chord progressions are going to be things that don’t necessarily pour out of me when I’m normally in my studio playing guitar but they’ll help me to build up my own riffs and string them together into an actual song. I could take the AI generated riffs and rewrite them in my own style and I could make massive changes to the arrangement of the track and it would be largely my own creation. But the basic fact remains that it’s the nature of the genre to be very generic. If I end up with a track that was co-created by AI and recorded by myself, containing no samples of audio from the AI, is it still my song? Is it derivative?
I love symphonic metal, and of course Nightwish, Therion, Blind Guardian are the masters (by the way, Xandria is very derivative already). But that doesn't mean that the genre cannot move forward. Think about symphonic classical music, and how it evolved in centuries, from Haydn to Beethoven to Wagner to Mahler to Shore. As a writer myself, I indeed aim to always add original ideas and create new song structures, and I think that I can do things that for AI would be utterly impossible. If you use AI, then you are condemned to be derivative, because AI has not really new ideas. It has no real thought nor deep understanding of music. So if you significantly modify AI's material, yes, you will have copyright, but I don't think that with AI you will ever be the new Nightwish. You will be just a fish swimming with other millions in the sea of cheapness.
Symphonic metal is exactly one area where I found that Suno is really good at, and I love it. One novelty is to pick different topics for the lyrics. For example, mixing symphonic metal with synthwave and science fiction lyrics.
@@NoidoDev I tested once Udio, and I obtained a pathetic version of an Epica song. Despicable. Yes, it sounded good, because it was stealing basically everything 😥 In my experience, if you are really expert in a genre, you can almost always locate the song from which AI takes “inspiration”, aka theft 😂
@@federicoaschieri For sure, but that’s my point. Even before AI there’s a ton of bands who are derivative (Xandria started out kinda doing their own thing and evolved into something more generic as they got more successful). So many bands are going to end up as that fish already and not be the next Nightwish but that’s fine because they just want to play a specific style of music that inspired them and not re-invent the wheel. And that’s what the AI is doing. So if my goal as an artist is to do something that’s already been done a million times and don’t care about being the best new thing then the AI can’t really hurt.
@@micah_noel I think there’s a difference. You can be in a genre, but also original, and not just taking a song from another band, and editing it until contentID doesn’t strike, as AI does 😂 For example Epica is a really great band and don’t copy Nightwish as I heard Xandria doing. If you want to be just a small fish, having fun, but with no chance of having a career, of course using AI is a good way to accompish that 😁
If necessary USIO and Suno should cut a deal with SoundCloud and train it’s AI on the material there. Most artists on SoundCloud would probably not be bothered either way or would take a small fee for it.
Why would they need to cut a deal when it's not even illegal? And it shouldn't be illegal. These big companies and some music VIPs want it that way, but it is in no one else's interest.
Meanwhile, the Majors are grappling with all this new fangled tech, the musical 'grass roots' fraternity get on with the job.
Embace away all they like, you can never replace the warmth and spontaneity of actual musicians, live or recorded.
As has been the case with evolution, what goes around comes around, and the same for this current wave.
Which, like all before, will reach a saturation point and therefore a stagnation stage, and the undercurrent 'grass-roots' will rise again
to the surface and the cycle will start all over. So, in this day and age, creators have to be more ... creative.
I hope they only have to pay for anything that is very close to some original. And the legacy industry has to provide some filter, or detection mechanism for that. Then they can fight and litigate over the exact borders.
Anything else should not fall under copyright of the legacy industry. Training is not sampling.
Suno and Udio should make their own cheap streaming services, where people can have songs created from time to time, if they sign up for a premium account. Otherwise listen to the music from the platform, but nothing too close to a legacy song.
I think that in order for rhe labels to prove the case..they have to prove they fed entire tracks into the system..burden isnt on suno or udio. I think its gonna be hard
it's cleary just user text that is assigned to a collection of indexed, single vocal samples. Then the vocals are autotuned and then tempo synced to fit on top of the backing track
I could do this without A.I with ease.
Then what the hell are you waiting for? Millions of dollars to be made..........
Regarding the producer tags... That is fascinating. Suno will be even more hard pressed to claim their AI didn't train on existing works. They're in a bind if they admit to training the AI on said works. They're similarly in a bind if they copped to having those very specific producer tags. But, the latter is possibly a worse position to defend from, since the AI still knows how to use them correctly, down to associated styles. So, in either case, the logical evidence is very convincingly showing that the AI _must_ have trained on this stuff.
Secondarily, I hope the lawsuit requires no more changes be made to the codebase, or a complete snapshot of the code and it's dependent assets be made. I'd assume as much, but I've been wrong often enough in the past.
You can upload your own audio to suno..how do we know sunos detractors dodnt load it ij themselves(
@@patrioticamerican9962 That's not just an unhelpful defense, it's irrelevant. They say it hasn't come into contact with copyrighted material. But, if they allowed someone to feed it individually, it's still come into contact with what they said it didn't. Now, they're absentee landlords, so to speak. And, really, defending from a position of "Well, *we* didn't feed it that stuff" requires borderline naivete. After all, the AI not only is using producer tags, but is matching the contextual use of them. It even knows basic styles to cross-associate the different names with. That would effectively require that every manual uploader feed the AI these specific elements in volume enough that it knew what producer tags go with a given style. And, again, Suno claimed very plainly that this hasn't happened. Even if they say "It want specifically is, your honor" it's still theirs, and they're still responsible for it.
I'm still waiting to hear the "new future music" these AI companies claim to be inventing out of nothing. Everything I've heard sounds like something I've already heard before and that's because it is something I've heard before just with different sounds and effects. The AI doesn't have a database of actual songs or even sounds for that matter, but there is a database nevertheless. That database is digital patterns based off of actual songs and sounds, then the AI just changes the "voice" of the pattern by overlaying other sounds over the digital pattern. Thus nothing entirely new is being made it is a digital pattern based on an existing song with a different sound overlay.
That A.I. business would run out of cash if they had to pay every time a copyright holder's music was used . I can safely assume they have limited resources .
Human artists would go broke if they had to pay royalties to the artists they listened to, every time they did a song.
The sampling problem is pissing me off immensely. I'm a producer. I like remixing. I can't put some of them remixes out as I have no way to get the license to the original. So if RIAA loses the lawsuit, we'll get into a phase where sampling is legally punished when you're human, but completely ok if you're an AI model. Come on. 😡
If I make and copyrighted a song and Ai copy’s it then that’s copyright infringement Period Sorry what’s next coping people and putting them in videos that they did not give permission to do so Do you see where this is going people
Parrots can reproduce sounds.
Sue them?
You cannot copyright nature.
A vibrating string, the sound of a drum, these things are elements of nature.
Colors cannot be copyrighted.
The english language cannot be copyrighted.
Artists are curators, picking and choosing elements that already exist and presenting them in new ways to convey new ideas.
Every chord progression and individual note already exist.
Musicians merely reorganize them.
You cannot blind the world to anything you make as it was all inspired by that which came before.
Every artist dips his cup into the well of history and nature.
The record companies would like to own the entire well.
This cannot be allowed to happen.
In a just and moral world, it would be impossible to steal anything with no physical weight.
I wouldn’t put it past record labels to create their own AI song creation tech. They own a ton of copyrights which means they have the right to use those copyrighted songs to train AI