Google DeepMind on March 25 released Lyria 3 and Lyria 3 Pro, its most capable music generation models to date. Lyria 3 (clip) generates up to 30-second tracks; Lyria 3 Pro produces full songs up to 3 minutes with structured composition — intros, verses, choruses, and bridges — and multilingual vocals. Both accept text and image prompts and embed SynthID watermarks in all output. The models are available via the Gemini API in paid preview, Google AI Studio, Vertex AI, and the Gemini app for paid subscribers.