Google puts its strongest music model inside Gemini

Google has added Lyria 3.5, its latest music-generation system, to Gemini, making song creation a built-in feature of the consumer AI app rather than a task requiring specialist music software. A user can describe a mood or genre—or supply a photo—then choose vocals or an instrumental track; Gemini can produce songs up to three minutes long, while developers can also access the model through Gemini’s API. Google says the update improves vocals and arrangements, but its larger significance is distribution: a chatbot used for writing and images is becoming a single creative workspace. Tracks made in the app carry Google’s invisible SynthID watermark, a nod to the attribution problems surrounding AI-made music.