YouTube is currently in the process of testing music-generating software that utilizes Lyria, an AI model developed by Google DeepMind.
The renowned video platform, Google, has introduced two new features, Desire Track and the Song AI Tool, to a limited group of testers. Desire Track is capable of transforming text-based prompts into short music samples inspired by artists like Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Sia, T-Pain, Troye Sivan, and Papoose. However, due to licensing constraints for educating Lyria on their music catalog, the selection of artists is currently restricted to this group.
One example of Desire Track’s output is a track titled “Dream Track’s Fast,” described as a “beautiful morning in Florida, R&B,” reminiscent of T-Pain’s style known for his autotuned vocals.
Moreover, YouTube is now allowing a limited number of creators to produce 30-second AI-generated music clips, suitable for posting as YouTube Shorts, which are typically one-minute videos.
On the other hand, the Audio AI Tool offers a more intriguing functionality, enabling users, especially those lacking musical instruments, to transform sound clips into different instrument sounds while retaining the original audio essence. An impressive demonstration showcases transforming simple vocal sounds into a complex orchestral score, resembling a movie soundtrack.
Currently, only a select group of musicians, songwriters, and producers have access to YouTube’s Audio AI Incubator system, empowering them to explore novel musical expressions without traditional instrument knowledge.
This innovative approach by YouTube aims to enhance artists’ creative processes and expand their artistic boundaries through AI-driven tools and experiences. The platform envisions closer interactions between fans and artists, fostering creativity and potentially influencing future software developments.
However, the challenges of creating AI-generated music that resonates authentically with audiences and navigating copyright issues remain significant hurdles. YouTube acknowledges these complexities and is actively engaging with the music industry to establish fair licensing agreements and copyright protection frameworks.
To address concerns about the potential misuse of AI-generated audio, Google DeepMind researchers have developed the SynthID tool integrated into the Lyria model. This tool embeds imperceptible watermarks into AI-generated tracks, allowing for the identification of artificially generated content. SynthID’s watermarking technique ensures the track’s detectability even after multiple alterations, safeguarding against deceptive practices.
DeepMind’s innovative approach not only enhances the authenticity of AI-generated music but also provides a mechanism for verifying the origin of audio content, contributing to a more transparent and accountable music creation process.
