Lyria 3.5 Launches in Google Flow Music With Better Vocals, Lyrics, and Tempo Control
Richer chord structures, more expressive vocals, and direct tempo controls - all landing in Flow Music today from Google DeepMind.
Lyria 3.5 is now live in Google Flow Music. Google DeepMind's updated music generation model brings four concrete improvements over the previous version: richer melodic structures, higher-quality lyrics with better prompt adherence, more expressive and emotionally realistic vocals with improved pronunciation, and direct controls over tempo and track duration. Flow Music is available at flowmusic.google.
Emotional Expression and Pronunciation Lead the Updates
AI music models have long struggled most with vocals. Suno and Udio both generate serviceable chord progressions and melody, but the vocal layer tends to sound robotic - mispronounced syllables, flat phrasing, no dynamic variation between the verse and the hook. Google says the new model addresses all three with more realistic emotional expression, improved pronunciation, and phrasing that shifts within a track rather than averaging to the same affectless center.
Tempo and Duration Controls Are New in This Version
Beyond vocals, Lyria 3.5 adds direct controls for tempo and track length - features that previously required workarounds or accepting whatever the model generated. Duration control is especially useful for creators building content around video timings or short-form formats where a 43-second track needs to land at exactly 43 seconds. Structural lyric awareness also improved. Verse and chorus sections now behave more distinctly, rather than producing loosely connected text set to a repeating beat.
Suno and Udio Hold Top Positions in AI Music Generation
Google launched three new Gemini models earlier this month, and Lyria 3.5 arrives as another update to a product track running separately from the company's core model releases. Suno and Udio lead AI music generation right now. Both offer free-tier access with no hard account barrier - making them easier starting points than Flow Music, which requires a Google account and keeps the experience inside a defined product shell. Meta's Muse generator occupies the same position in AI image creation: Google and Meta are both building creative AI tools aimed at markets that smaller, focused startups currently own.
No API or Pricing Published Outside Flow Music
Lyria 3.5 has no separately published pricing. Google has not confirmed whether the model will power music generation elsewhere in the product range - Gemini features, YouTube Creator tools, or Google Workspace. Whether vocal improvements hold up across full production sessions rather than short demos is the real question, and that answer comes from musicians running full tracks, not from a four-bullet launch post.