Google DeepMind has shipped Lyria 3.5, its latest music generation model, now live inside Google Flow Music. The update targets four specific weaknesses of prior versions: musicality, lyric coherence, vocal quality, and user control over creative output. These are not incremental tweaks. They represent the core failure points that made earlier AI music tools sound like demos rather than finished work.
The reason to read the full piece is not the launch announcement. It is the specifics of how DeepMind approached vocal generation and creative control, two problems that have consistently broken competing models from Suno and Udio. The technical decisions behind lyric coherence alone reveal meaningful choices about how the model handles timing, syllable stress, and melodic phrasing simultaneously.
What comes next is the real question. Lyria 3.5 sits inside Flow Music, a product aimed at professional and semi-professional creators. Google is positioning this as infrastructure for human-led production, not a replacement workflow. Whether the creative control features are deep enough to hold that position against purpose-built tools will become clear fast.
[READ ORIGINAL →]