← Back to stories Modern home studio setup with music production equipment including keyboard, computer, and speakers.
Photo by Jan Makwela on Pexels
凤凰科技 2026-03-25

Google’s Lyria 3 Pro pushes AI music from snippets to full songs — and embeds provenance by design

What was announced

Google (谷歌) this week unveiled Lyria 3 Pro, an upgraded music-generation model that extends output length from the roughly 30‑second clips of Lyria 3 to full pieces up to three minutes. Short and startlingly polished music snippets are no longer the limit. Google said Lyria 3 Pro also improves structural understanding — users can now prompt for an intro, verse, chorus and bridge — and steer composition with far finer control.

Where it will appear

Lyria 3 Pro will be integrated into Gemini and Google Vids, and reportedly will be available in the Gemini app only to paid subscribers. It will also be added to tools for creators: Google plans to fold the model into Vertex AI (currently in public preview), Gemini API and AI Studio so developers and enterprises can embed music generation into apps, games and branded audio experiences. It has been reported that Google last month acquired ProducerAI, a generative‑music tool, and will tie that capability into the Lyria stack.

Data, rights and provenance

Google says Lyria 3 Pro was trained on partner datasets and “licensable” content drawn from Google and YouTube, and the company insists the model will not directly mimic specific artists — although it may produce music “inspired” by a named artist if prompted. To help platforms and regulators identify AI output, Google will embed SynthID markers in all tracks produced by Lyria 3 and Lyria 3 Pro, a technical provenance signal designed to make generated audio discoverable and traceable.

Industry and regulatory context

Why does this matter beyond Google’s product roadmap? Streaming platforms and rights holders are scrambling to set guardrails. This week Spotify and Deezer rolled out tools to help detect or remove AI‑generated tracks falsely attributed to real artists. Regulators in the U.S., EU and China are increasingly focused on deep‑synthesis and copyright issues, and provenance mechanisms like SynthID will likely become a table‑stakes feature for major vendors. The race is on: can platforms balance rapid creative tooling with artist rights and regulatory scrutiny?

AI
View original source →