Google DeepMind has unveiled significant updates to their AI music creation tools, focusing on three main platforms: MusicFX DJ, Music AI Sandbox, and YouTube Shorts' Dream Track.
The centerpiece of this release is MusicFX DJ, which has been reimagined with expanded controls and improved audio quality. This tool generates new music in real-time by allowing users to mix musical concepts through text prompts, unlike traditional DJ tools that blend existing tracks.
The technology employs two key technical innovations:
- An offline generative music model adapted for real-time streaming.
- A system that combines multiple text prompt embeddings to guide the music's style.
The platform now outputs professional-grade 48 kHz stereo audio and enables users to download 60-second clips of their creations.
Working with six-time GRAMMY winner Jacob Collier, the team developed intuitive controls that let users manipulate instrumentation, create breakdowns, adjust textural aspects, and control key and tempo. Users can now share sessions with others, who can watch performances and take over the controls mid-session.
The Music AI Sandbox, while not publicly available, serves as a testing ground for professional musicians and producers. It's being updated with multi-track views and precise controls for composition refinement. Meanwhile, YouTube's Dream Track experiment has evolved to offer U.S. creators the ability to generate instrumental soundtracks using text prompts.
To ensure responsible deployment, all music generated through these platforms is watermarked using SynthID. This initiative represents Google DeepMind's ongoing collaboration with the music industry to develop tools that serve both professional musicians and newcomers to music creation.
