YouTube creators may be able to upload music that sounds like it was actually sang by famous musicians one day. According to Billboard and Bloomberg, the video hosting platform is developing an artificial intelligence tool that can mimic the voice of recording artists. It was also hoping to introduce the feature during its Made On YouTube event in September, where it debuted other AI-powered creator tools including a generative green screen. Obviously, its plan didn't push through, and it was apparently because negotiating with recording companies is taking time. The technology is new, after all, and there are a lot of questions that need to be answered first.
Billboard says one of the key issues they're still hashing out is how monetization would work. Will artists be paid for the music YouTube feeds its AI model, or for the output made using the AI tool? Further, how would songwriters be compensated for any of their music used to train YouTube's AI? Recording companies still reportedly have questions about how the AI model is trained, as well, and about how artists will have the power to opt in or out. If and when the tool does come out, YouTube intends to beta test it first by giving a select group of creators access to the voices of artists who opt in.
While negotiations are taking time, recording companies are receptive to YouTube's idea, according to the sources. That's because they're looking at the use of AI in music as an inevitability, and they're concerned about getting left behind if they resist licensing deals like what YouTube is offering. As Bloomberg notes, though, YouTube is walking a tricky path and may have to brace itself for legal issues in its quest to explore the use of generative AI in music. OpenAI, for instance, had been sued by a group of authors, including John Grisham and George R.R. Martin, for the alleged use of their novels to train its large language model.