Google DeepMind’s latest music model landed today, July 29, 2026, and it’s arriving with a promise that has eluded most AI composers so far: songs that sound like they were made to move you, not just to prove a machine can string notes together. Lyria 3.5 is rolling out inside Google Flow Music, and the headline change is emotional nuance.
Where earlier generative music tools tended to produce competent but soulless loops, Lyria 3.5 focuses on the parts humans actually care about. Google says the model builds richer, more complex melodic structures and generates higher-quality lyrics with better prompt adherence and a stronger sense of song structure. In plain terms: verses, choruses and bridges that hang together, instead of a wall of pleasant-sounding mush.
The vocals get the biggest glow-up. Lyria 3.5 delivers more realistic and emotionally nuanced vocals with improved pronunciation — the kind of detail that separates a convincing performance from an uncanny one. If a robot singer flubs its consonants, the illusion collapses fast, so tightening enunciation is a smart place to spend the compute.
Crucially, this isn’t a black box that spits out a finished track and shrugs. Google has leaned into creative control, handing users direct levers over the mix:
- Tempo — set the pace of the piece
- Vocals — shape the sung performance
- Drums — dial in the rhythm section
- Bass — control the low end
- Track length — anything from 30 seconds to three minutes
That range is telling. Thirty seconds is jingle and social-clip territory; three minutes is a full radio-length song. Lyria 3.5 is clearly aiming to serve both the hobbyist knocking out a background loop and the creator who wants a complete track with a real arc.
Access comes bundled with Google’s AI subscription tiers rather than as a standalone product. Google AI Plus members get the benefits of Google Flow Music’s Starter plan, AI Pro members get the Plus plan, and AI Ultra members get the Member plan. In other words, the more you’re paying Google for AI, the more Flow Music muscle you unlock. Google hasn’t published separate pricing for Lyria 3.5 itself.
The bigger picture is that generative music is maturing past novelty. The first wave of these tools was about proving feasibility — yes, an AI can make a song. Lyria 3.5’s pitch is about taste and steerability, giving people the knobs to push a track toward what they actually hear in their heads. Whether that lands with musicians or unnerves them, the tooling is getting more serious, and this release is a clear step up the ladder.
