Google rolls out Lyria 3.5 for more expressive AI music creation

Google has launched Lyria 3.5, a new version of its music-generation model, in Google Flow Music. The update is aimed at helping users produce fuller songs while giving them more control over the creative process.

According to Google, Lyria 3.5 can create richer and more complex melodies designed to sound more natural. The model also improves lyric generation, with better adherence to prompts and a stronger understanding of song structure. That could make it easier for creators to request specific themes, sections, or lyrical formats without as much manual rewriting.

Vocals receive a major upgrade as well. Google says the new system produces performances with greater emotional range and realism, alongside improved pronunciation. For AI music builders, those changes could make generated tracks more usable as demos, soundtrack elements, or starting points for further production.

The release also adds more direct control over tempo and output length. These controls matter in practical workflows: creators can better match a generated song to a video, prototype a particular arrangement, or iterate within a target duration rather than repeatedly regenerating tracks and editing them afterward.

Lyria 3.5 is available now through Flow Music. Google’s announcement focuses on the model’s capabilities and does not describe a separate developer API or standalone model release, so the immediate opportunity is centered on experimentation inside the Flow Music experience.

For people building with generative audio tools, the update reflects a broader shift from one-click song generation toward systems that offer more dependable control over composition, performance, and structure.

Source: Google DeepMind Blog