Hip-hop mogul Jay-Z has become an unlikely source of inspiration in the burgeoning artificial intelligence identity theft community known as deepfakes and the record label. the rap icon took note and issued copyright notices to close some of it.
It’s not that Jay-Z is slandered without his consent, as the majority of deepfakes tend to do by sticking the heads of female celebrities and other women on the bodies of adult film actresses. Rather, it’s about the world of safe, work-only deepfakes, audio only that are mostly tried out in parody, as well as showing the impressive technical sophistication that the field has produced in just a few short years.
In a fascinating deep dive from XOXO festival co-founder Andy Baio on his website Waxy, Baio delves into AI-based imitations of Jay-Z on YouTube, in particular a creator using Jay-Z’s iconic voice and hip-hop streams to rhyme classics such as the solo “To Be or Not to Be “by William Shakespeare Hamlet and “We didn’t light the fire” by Billy Joel. “
In a strange turn of events, Roc Nation LLC, Jay-Z’s full-service entertainment agency, has filed copyright strikes against YouTube downloads of the aforementioned deepfakes. Reviews specifically cite AI, with Roc Nation writing, “This content illegally uses AI to usurp the identity of our client’s voice,” according to Baio’s conversations with the creator, who remains anonymous and passes by the Voice Synthesis online handle. The channel itself has nearly 40,000 subscribers, and many of its videos have garnered hundreds of thousands of views.
Curiously, Jay-Z’s deepfake videos featuring the rapper’s synthetic voice rhyming the book of genesis and the famous Navy Seal meme copypasta remain on YouTube. But you can’t attribute that to those who are free to use; “We did not light the fire” is of course copyrighted, but Shakespeare’s works are in the public domain.
All Voice Synethesis videos are created by feeding Google’s open source Tacotron 2 text-to-speech model with songs and lyrics from Jay-Z and having the synthetic voice of the pre-written text read. The situation raises fascinating questions about what exactly is violated here if the synthetic voice simply produces original content using the likeness of a celebrity. For a closer look at the copyright issues at stake here, I strongly suggest that you read Baio’s analysis in its entirety, as it goes into the finer details of fair dealing and why Jay’s claims – Z may not hold out.
“It seems reasonable to assume that a model and sound generated from copyrighted audio recordings would be considered derivative works. But is this a copyright infringement? Like almost everything in the copyright world, it depends – how it was used and for what purpose, “says Baio. He uses a record producer example starring Jay-Z on a new single without his permission as an obvious legal transgression before arguing why Vocal Synthesis could possibly be right here.
The Deepfakes are on cloudy legal ground. Some US states like California have banned political candidates from broadcasting deepfakes in an attempt to influence an election, while Virginia last year expanded its revenge pornography laws to cover computer-generated images and videos. A number of social platforms – including Facebook, Reddit and Twitter – all have deep bans on books that cover a wide variety of content, in most cases to prohibit manipulated or misleading video or audio that is designed to cause harm.
But none of these bans appear to cover harmless entertainment, to which the Voice Synthesis videos appear to fall. “I was quite surprised to receive the withdrawal order. As far as I know, it was the first time that YouTube deleted a video to impersonate an AI voice, “said the account creator to Baio in an interview, which you should definitely read in whole on his website. “I have been posting these types of videos for months and have not deleted other videos for this reason. There are also several other channels producing text-to-speech videos similar to mine, and I am not aware of any that have deleted videos for this reason. ”
Speech synthesis has a point; AI-based voice spoofing is everywhere on the Internet, especially on YouTube. It’s just that, in this particular case, a well-known celebrity noticed when it became the subject and took action that could have far-reaching consequences if it became standard practice to moderate these types of deepfakes in the future. . .
“I’m not a lawyer and I haven’t studied intellectual property law, but logically, I don’t really understand why imitating a celebrity’s voice using an AI model should be treated differently from someone who naturally makes an (extremely precise) impression of this celebrity’s voice. “, Argue Voice Synthesis.
YouTube did not immediately respond to a request for comment. Roc Nation does not list telephone numbers or email addresses on its website to contact the company or its label subsidiary.