Google releases Lyria 3 Pro music-generation model, reportedly able to produce three‑minute tracks

What Google announced

Google has unveiled Lyria 3 Pro, a new iteration of its Lyria family of music-generation models that, it has been reported, can generate coherent three‑minute music tracks. The company says the model expands on earlier short‑form capabilities to deliver longer, stylistically controlled compositions — from instrumental pieces to fuller arrangements — aimed at creators, game developers and music services. Reportedly the launch includes APIs and tooling to let developers integrate longer-form generated music into apps and workflows.

How it works and the limits

Google has not published full technical details, but Lyria 3 Pro is reportedly tuned for longer temporal coherence and better instrumental separation than previous models, and includes safety filters and export controls designed to reduce misuse and copyright infringement. How the model was trained — and on which corpora — remains unclear; those details matter for licensing and rights questions, and have not been fully disclosed. Will automated music at this length satisfy composers, or merely sharpen existing legal disputes about training data and ownership?

Why it matters for China and the rest of the world

For Western readers new to China’s tech scene: Chinese tech giants are already advancing generative audio and music capabilities. Baidu (百度), ByteDance (字节跳动) and Tencent (腾讯) have active AI research and commercial offerings that could compete with or adapt Google’s approach for domestic platforms. Geopolitical context is relevant too — AI models and cross‑border data flows sit in the middle of trade tensions and export‑control debates, and regulators in multiple jurisdictions are watching how large models handle copyrighted material.

What comes next

Google’s rollout raises practical and policy questions: how will streaming platforms, rights holders and regulators respond? It has been reported that Google plans commercial licensing options and content‑management tools, but adoption will hinge on transparency about training data, provenance controls, and enforceable safeguards. For creators, longer‑form generation opens new possibilities; for the music industry, it intensifies the conversation about how to balance innovation with fair compensation and copyright protections.