Google updated its frontier music model Lyria 3.5 within the Flow Music platform, adding tempo control tools and the ability to export individual tracks, which moves audio generation from the realm of entertainment to the category of professional tools.
What happened
The updated Lyria 3.5 model offers more dynamic and expressive vocals, improved musicality of arrangements, and more accurate adherence to text prompts. Key technical innovations were the implementation of tempo (BPM) control and support for exporting separate tracks (stems) for subsequent editing.
Context
The development of Lyria 3.5 marks a transition from a "black box" model that outputs a finished audio file to controllable synthesis. This allows integrating the capabilities of generative AI directly into existing music production workflows.
Why this matters for the industry
Improving the quality of vocals and instruments brings AI music closer to commercial standards. Support for stem export makes the technology suitable for professional workflows, creating a foundation for the emergence of new plugins and deep integration of generative models into professional software (DAW), such as Ableton or Logic Pro.
Why this matters for users
Musicians and producers get a tool for rapid prototyping of high-quality audio elements. The ability to generate individual layers, for example, only a vocal part or a bass line, allows using Lyria 3.5 as an advanced assistant within familiar digital audio workstations (DAW).
What is still unknown / limitations
At the moment, technical parameters for full industrial integration remain undefined, such as the availability of a stable API, latency, and cost of use (pricing). There are also legal risks in the area of intellectual property and the creation of derivative works.
Sources
Author
Look at AI, editorial team