Testing the Dream Track Alongside Various Songs Using AI Tools
By Jon Porter, a seasoned writer with expertise in consumer software releases, EU software plans, online platforms, and electrical keyboards.
Google is currently in the process of testing innovative AI capabilities for YouTube, enabling users to create music tracks solely based on text prompts or hummed tunes. One of these features, Dream Track, has been introduced to a select group of creators on the platform. This tool is designed to automatically produce short 30-second music tracks inspired by popular artists. Dream Track involves collaboration with nine different musicians who are working closely with YouTube on this project. Additionally, YouTube is showcasing cutting-edge technology that can generate music tracks from simple hummed melodies.
This development comes as YouTube navigates the evolving landscape of standards and copyright regulations concerning AI-generated music while also maintaining its relationships with major music labels. The issue gained significant attention when an AI-generated “Drake” track went viral earlier this year, prompting YouTube to partner with Universal Music to establish guidelines for AI-generated music on the platform.
The Dream Track feature is currently undergoing testing with a limited group of creators in the US and can create tracks inspired by nine artists: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan. Through live demonstrations, users were shown how a prompt like “A ballad about how opposites attract, with a cheerful sound” could generate a track in the style of Charlie Puth, or how “A beautiful morning in Florida, R&B” could be used to create a T-Pain song. This tool can produce complete songs, backing tracks, and AI-generated lyrics tailored to the artist’s style.
The intention is to integrate these tracks into YouTube’s Short service, akin to TikTok. This is where YouTube recently introduced another AI tool called Dream Screen, allowing users to create videos and images for use as backgrounds.
In addition to Dream Track, YouTube has unveiled other Music AI Tools that enable users to create audio tracks without needing to play traditional instruments. Demonstrations have shown how the tool can transform a whispered melody and a text prompt like “saxophone solo” into a saxophone track. Other examples include beatboxing converted into a drum circle, vocals transformed into instrumental compositions, and melodies from an electronic keyboard reimagined as vocal performances.
Participants in YouTube’s Audio AI experiment will have the opportunity to explore these tools starting later this month.
These new AI tools are powered by the Lyria song generation model developed by Google’s DeepMind. According to a blog post from DeepMind, tracks created using Lyria will be tagged with a SynthID logo that is imperceptible to the naked ear but remains detectable even after modifications. This ensures that the presence of AI-generated content from Lyria can be verified, even if the track is altered, compressed, or sped up.
The release of these tools coincides with YouTube’s announcement of updated content guidelines for AI-generated deepfakes, aimed at safeguarding users and the interests of the music industry partners.
