Google DeepMind launched Lyria 3.5 on July 29, 2026, rolling out its newest music generation model directly inside Google Flow Music. The deployment delivers specific advancements across musicality, lyrics, vocal quality, and creative control, with the stated goal of empowering creators to craft richer tracks and produce songs they love.
Musical composition receives significant updates under the new architecture. Google DeepMind reported that Lyria 3.5 creates richer, more complex melodic structures designed to sound more natural, improving the quality and progression of melodic elements across generated tracks.
For written composition, the updated model generates higher quality lyrics with improved prompt adherence and enhanced structural awareness. These refinements allow the generation system to follow user prompt instructions more closely while maintaining organized song structure across different sections.
Vocal performance features multiple technical updates in the new release. Google DeepMind noted that Lyria 3.5 brings more expression and emotion to generated songs through more realistic and emotionally nuanced vocals. The model also incorporates improved pronunciation across vocal lines.
The update expands operational parameters by giving users more direct creative control within Google Flow Music. Creators can more easily control output tempo and specify output duration. Google DeepMind made Lyria 3.5 available for immediate use inside Flow Music on the day of the announcement.
