Google has released Lyria 3.5, an upgraded music generation model integrated into Google Flow Music. The new version generates complete tracks between 30 seconds and 3 minutes in length.
The key addition is "Selective Section Painting," a feature that lets users edit specific sections of a generated track without regenerating the entire piece. This targeted editing capability addresses a major workflow friction point in AI music generation. Previously, users typically had to restart from scratch if they wanted to modify a verse, chorus, or instrumental section.
The feature works within Google Flow Music's interface, allowing granular control over which portions of a track users want to adjust. This positions Lyria 3.5 as more practical for iterative music creation compared to earlier generation-only systems.
Google has not disclosed training data details for Lyria 3.5, continuing a pattern of opacity around its music models. The company has faced ongoing questions about how its generative audio systems were trained, particularly regarding whether copyrighted music was included without explicit licensing or artist consent.
Lyria 3.5 joins a growing field of AI music generation tools. Competitors including Meta's MusicGen, OpenAI's Jukebox, and specialized startups have pushed capabilities forward significantly. The introduction of section-level editing represents incremental progress on usability rather than a fundamental architectural breakthrough.
Google's release of Lyria 3.5 reflects broader industry momentum in AI music generation moving from simple prompt-to-audio tools toward systems with finer creative control. The selective editing approach gives musicians and producers more practical leverage in leveraging AI for composition and arrangement work. However, the lack of transparency around training data remains a substantial concern for creators worried about whether their work contributed to the model's training without compensation or consent.