Google just made its latest music generation model available to developers. Lyria 3 launched today in paid preview through the Gemini API, with free testing available in Google AI Studio.
The details are thin so far. Google's announcement, written by Alisa Fortin, confirms the model exists and that developers can start building with it, but doesn't share much else. No word on supported audio formats, generation length limits, quality benchmarks against competitors, or how it improves over previous Lyria models. Pricing for the paid preview tier hasn't been disclosed either.
What We Actually Know
Lyria 3 is designed for developers who want to add AI music creation to their apps and products. Two access paths are available right now:
- Gemini API (paid preview): For production-grade integration into apps
- Google AI Studio: For free testing and experimentation
That's about it. Google is clearly positioning this as infrastructure for other developers to build on, not a consumer-facing music tool. Think background music generators, game soundtrack tools, or content creation platforms that need royalty-free audio.
The Competitive Picture
The timing is notable. Music generation has been one of the quieter corners of generative AI compared to text, image, and video. Suno and Udio have built consumer followings, but neither has the kind of API-first developer platform that Google can offer through its existing Gemini infrastructure. By plugging Lyria 3 into the same API developers already use for text and image generation, Google is betting on convenience and ecosystem lock-in.
The missing details matter, though. Developers evaluating music generation APIs care about output quality, latency, cost per generation, and content moderation policies. Google hasn't addressed any of these yet. Until the documentation fills in those gaps, this is more of a signal of intent than a product you can seriously evaluate.