Lyria 3.5 brings lyrics, vocals, and tempo control into Flow Music

The next fight in AI music is not only whether a model can generate a convincing track. It is whether creators can keep shaping that track after the first output. Google said on July 29, 2026 that Lyria 3.5 is rolling out in Google Flow Music with improvements across musicality, lyrics, vocals, and creative control. The model is available to try in Flow Music now.

The most practical change is control. Google says Lyria 3.5 makes it easier to adjust tempo and duration, two settings that matter when a generated song needs to fit a real creative brief rather than stand alone as a short demo. Music tools often impress on first generation, but production work depends on revision: making a section longer, slowing a track down, tightening pronunciation, or changing the shape of a vocal line. Lyria 3.5 moves closer to that iterative workflow.

Google describes three quality improvements around the controls. Improved musicality is aimed at richer and more complex melodic structures that sound more natural. Enhanced lyrics are meant to improve prompt adherence and structural awareness, reducing the gap between what a user asks for and what appears in the song. Improved vocals focus on expression, emotional nuance, realism, and pronunciation. The announcement does not include public benchmark numbers or third-party listening tests, so the claims will need to be judged through actual use.

The product context makes the launch more interesting. Google has been expanding creative AI through Flow and its video models, and adding Lyria 3.5 to Flow Music points toward a more connected production surface for image, video, and audio generation. The unresolved questions are still significant: copyright, voice likeness, dataset transparency, and how musicians will be credited or compensated. But for users, the concrete shift is clear. AI music is becoming less like selecting one generated sample and more like editing a track through repeated, controllable changes.