Alphabet rolls out Lyria 3.5, boosting music generation in Google Flow Music

Alphabet is updating its music-generation offering with a new model it says raises the quality of generated songs and gives users more steering over the output. In a post published July 29, the company said it is rolling out Lyria 3.5 in Google Flow Music, positioning the update around “musicality, lyrics, vocals, and creative control.”

The company’s main claims are focused on output quality. Alphabet said Lyria 3.5 improves “musicality,” including the ability to produce richer, more complex melodic structures that sound more natural. It also said the model improves lyric generation, citing higher-quality lyrics with better prompt adherence and structural awareness.

On vocals, Alphabet said Lyria 3.5 brings more expression and emotion to songs and uses more realistic, emotionally nuanced vocals. The update also includes improved pronunciation, according to the company, an area that can matter for music systems that attempt to align lyrics to sung phrasing.

Beyond quality, Alphabet highlighted creative controls. The company said Lyria 3.5 makes it easier for users to control tempo and the duration of outputs, aiming to reduce the “randomness” users often associate with generative systems. Alphabet did not provide technical details about how those controls are implemented, only that the experience is improved.

Google Flow Music is the consumer-facing interface where the new model is available. Alphabet’s post describes the goal as helping users create songs they “love,” with creative control, and it directs readers to try Lyria 3.5 in Flow Music.

For Alphabet, music generation is part of a broader push to commercialize AI capabilities through user-facing products. The company’s language suggests it is treating model upgrades as product iterations, where improvements in output quality and controllability can translate into higher user satisfaction and retention in an interactive tool.

The update also reflects a familiar pattern in AI media systems: companies tend to refine generation in stages, first targeting what users can perceive most immediately, such as melody, lyric coherence, and vocal expressiveness, and then layering in controls that help users shape outcomes rather than simply request them.

Still, the company’s announcement leaves several questions unanswered. It did not disclose benchmark scores, examples, or comparative performance versus prior versions of Lyria, nor did it specify which parts of the pipeline are improved. It also did not detail how creative-control features interact with different prompts, languages, or genres.

What to watch next is whether Alphabet expands Lyria 3.5’s availability beyond the Flow Music experience and how the company measures user impact. If the model’s claims around prompt adherence, lyric structure, and controllable tempo and duration prove out in everyday use, it could set a higher bar for generative music tools, where users increasingly expect both quality and steering.