For paternal videos, YouTube has partnered with musicians such as John Legend and Sia to offer AI-generated renditions of their vocal styles.
These exclusive 30-second segments are being created through a collaboration between YouTube, owned by Google, and a specialized AI unit within the company’s music generation framework. The roster of artists involved in this innovative project includes Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Sia, T-Pain, Troye Sivan, and Papoose.
YouTube has introduced a trial initiative named Dream Track, which is currently accessible to a select group of content creators in the United States utilizing the Shorts feature, akin to TikTok.
As per a recent blog post on YouTube, composers can craft a complete 30-second melody by merely inputting text into the designated field.
Lyor Cohen and Toni Reid, YouTube executives, explained, “By entering a concept into the creative prompt and selecting a featured artist from the carousel, creators can instantly generate a custom Shorts soundtrack showcasing the AI-rendered voice of the chosen artist for their Short video.”
Examples shared by YouTube online demonstrate snippets of the AI-generated Charlie Puth singing “Baby We’ve Got Nothing in Common,” prompted by a request for “a ballad on the theme of opposites attracting, with a lively acoustic vibe.” Similarly, T-Pain’s rendition is inspired by the prompt “An enchanting morning in Florida, R&B,” capturing a narrative of awakening in a “state of light” in his signature autotuned style.
The Dream Track experiment encompasses AI-generated lyrics, melodies, and phrases. Approximately 100 Shorts creators and artists in the US currently have access to this feature on YouTube. Once the track is generated, users can leverage functions like “use this audio” and “edit this clip” to incorporate and remix it as desired.
YouTube aims to explore additional AI-generated tools later this year, such as transforming a hummed tune into an electric guitar riff or reimagining a classic pop track into a reggaeton composition. Powering the Dream Track trial is Lyria, a music-generation AI model developed by Google DeepMind, the tech giant’s AI division headquartered in London.
YouTube also announced a forthcoming tool that will empower music labels and distributors to address unauthorized AI-generated content mimicking an artist’s voice. While whimsical content and deepfakes of public figures may persist on the platform, users will have the ability to report such instances.
In light of ongoing disputes over the unauthorized use of artists’ voices and works in AI-generated content, the Dream Track project emerges. The unauthorized release of a track titled “Heart on My Sleeve,” featuring AI-generated lyrics purportedly in the style of Drake and The Weeknd, underscored the growing prominence of artificial music in the industry this year.
Amidst legal battles, music labels like UMG are pursuing legal action against AI entities like Anthropic for allegedly replicating copyrighted lyrics from renowned songs by artists like Gloria Gaynor and Katy Perry. Anthropic has yet to provide a response to these allegations.
Furthermore, a significant development occurred when the head of audio at Stability AI resigned due to ethical concerns regarding the company’s stance on the “fair use” defense for utilizing copyrighted material in training AI models without compensating rights holders.
Ed Newton-Rex, in a blog post on X (formerly Twitter), criticized the practice of AI companies leveraging copyrighted content to train models that could potentially compete with original works. He emphasized the importance of respecting copyright laws in a creative industry where creators heavily rely on these protections.
Newton-Rex highlighted his commitment to developing music-generating models based on authorized data and has reached out to Stability AI for further insights on this matter.
