In the rising world of content creation, Google's subsidiary, YouTube, unveils an experimental AI tool designed to replicate the singing prowess of renowned artists, with their explicit consent. This endeavor encompasses the trial of the Dream Track feature, alongside an array of other AI-driven musical tools capable of fabricating instrumental tracks derived from textual cues and melodic hums.
Google is presently scrutinizing novel generative AI functionalities for YouTube, empowering users to compose musical tracks using either a simple textual prompt or a hummed melody. The inaugural feature, Dream Track, discreetly disseminated among select platform creators, autonomously crafts succinct 30-second musical compositions mirroring the signature styles of eminent artists. Nine artists, in collaboration with YouTube, are part of this venture, where the platform addresses the intricacies of AI-generated music, ensuring adherence to emerging norms and copyright regulations while preserving amicable relations with major music labels.
YouTube introduces Dream Track to a limited cohort of U.S. creators, offering the capacity to emulate the musical styles of nine diverse artists: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan. Through illustrative video demonstrations, the platform showcases how a prompt such as "A ballad about the magnetic allure of opposites, exuding an upbeat acoustic vibe" can spawn a composition akin to Charlie Puth's oeuvre. Similarly, a prompt like "A radiant morning in Florida, infused with R&B vibes" translates into a T-Pain-inspired melody. The software seamlessly generates lyrics, a backing track, and an AI-rendered voice, meticulously emulating the chosen artist's style.
These crafted tracks find their purpose within YouTube's Shorts service, akin to TikTok's format. In tandem, YouTube introduces another AI feature, Dream Screen, dedicated to generating captivating videos and photos suitable for backgrounds, as revealed in September.
Beyond the Dream Track revelation, YouTube unveils additional Music AI Tools, designed to facilitate the construction of musical compositions without the need for conventional instruments. Demonstrated through video presentations, the tool showcases the creation of a saxophone track by amalgamating a hummed melody with the textual cue "saxophone solo." Other instances exhibit beatboxing metamorphosing into a drum loop, vocalization evolving into an orchestral score, and chords from a MIDI keyboard transforming into a vocal choir.
Participants in YouTube's Music AI incubator will gain access to these innovative tools later in the year. The underlying technology driving these AI tools is Lyria, a music generation model developed by Google's DeepMind. According to an accompanying blog post from DeepMind, tracks produced using Lyria will bear a SynthID watermark, inaudible to the naked ear, ensuring traceability even after modifications, such as added noise, compression into an MP3 file, or alterations in playback speed.
This unveiling follows YouTube's recent announcement of new content guidelines to counter AI-generated deepfakes, a measure intended to safeguard individuals and, by extension, the platform's alliances within the music industry.