Innovation

YouTube Will Let You Make Songs With AI Clones of Singers

OFF KEY

Writing music might get a lot easier—and maybe a lot worse—as the platform releases Dream Tracks today to a small group of creators on YouTube Shorts.

In this photo illustration, the American video-sharing website platform owned by Google, YouTube, logo seen displayed on a smartphone with an Artificial intelligence (AI) chip and symbol in the background.
SOPA Images / Getty Images

Making music is about to get a whole lot easier—and potentially a lot worse too. YouTube announced Thursday that it’ll be releasing a product using artificial intelligence that allows users to create songs with just a prompt.

The tool is called Dream Tracks and will be available to users to create 30-second tracks for YouTube Shorts, the platform’s TikTok competitor. The AI generator allows users to create music using deepfaked voice clones of nine famous singers including John Legend, T-Pain, Sia, Demi Lovato, and Charlie XCX.

YouTube said in a blog post that all of the artists have consented to their voices being used, and added that it’ll be releasing Dream Tracks today to “a small group of artists and creators,” before rolling it out to the “broader music community” at an unspecified date in the future.

ADVERTISEMENT

“At this initial phase, the experiment is designed to help explore how the technology could be used to create deeper connections between artists and creators, and ultimately, their fans,” the blog post said. It later added that music can be created by “simply typing an idea into the creation prompt and selecting a participating artist that appears in the carousel.”

This is a part of YouTube’s broader strategy of collaborating with singers and songwriters as a part of their Music AI Incubator to “explore the potential of AI features to help artists and creators stretch their imaginations and augment their creative processes.”

To that end, the platform also announced that it’s working on other music creation tools set to be released later this year such as a feature that allows musicians to upload files of their songs and augment them to sound like different genres. Users can also create entirely new songs by simply humming a melody and prompting the AI to turn it into different instruments.

In demo videos, the tool is used to do things like turning a single voice humming into an orchestra and a person beatboxing into a drum beat. Each song generated using AI will include a digital watermark created by Google DeepMind that’s undetectable to the naked ear.

The launch comes on the heels of the platform’s announcement on Tuesday that it’d be cracking down on videos containing AI-generated songs that contain the voice of another artist without their consent. The policy allows music labels to flag offending videos to be taken down for copyright infringement.

The latest announcement also highlights the tightrope that the platform is attempting to walk when it comes to AI-generated content, which is coming under increasing scrutiny from lawmakers, music publishers, and users at large. While the company has shown that it wants to crack down on non-consensual usage such as the viral AI-generated Drake song, it’s also attempting to leverage the tool in order to draw users to its products like YouTube Shorts.

The tool represents another threat to musicians and creatives who have been sounding the alarm on the risks of generative AI to their livelihoods since ChatGPT’s release one year ago. While Dream Tracks seems fairly limited for now, it’s only a matter of time before these types of technologies become more available—and that’s a song and dance a lot of musicians don’t want to see.

Got a tip? Send it to The Daily Beast here.