Google Integrates AI Music Model Into Gemini
Google has launched Lyria 3, a new AI music generation model, and integrated it into its Gemini platform. The tool allows creatives to generate short songs up to 30 seconds long using text, photo, or video inputs. This development enables new cross-modal workflows for multimedia storytelling and automated soundtrack creation.
- To distinguish between different use cases, Google has structured Lyria into three distinct products: Lyria 3 is the consumer model integrated into the Gemini app for generating 30-second tracks with vocals and cover art; Lyria 2 is an instrumental-only developer API available via Vertex AI with more technical controls like negative prompts; and Lyria RealTime is an experimental API for live, interactive music creation. - Every audio track generated by Lyria is watermarked using Google's SynthID technology. This imperceptible watermark is designed to be detectable even after compression or speed changes, allowing anyone to upload an audio file to Gemini to verify if it was created by Google's AI. - The model generates complete, multi-instrument arrangements from scratch at 48kHz stereo quality, simultaneously handling melody, harmony, rhythm, and vocals rather than stitching together pre-made loops. - The question of authorship for AI-assisted music is being actively debated, with the U.S. Copyright Office ruling in early 2025 that AI-generated work can only be copyrighted if it includes "meaningful human authorship." Works created solely from a text prompt without further creative human intervention fall into the public domain. - A specific application of Lyria is "Dream Track," an experimental tool in YouTube Shorts that allows creators to produce 30-second soundtracks for their videos using text prompts. - For more interactive and collaborative workflows, the Lyria RealTime API is designed to be controlled by a user on the fly, allowing for the blending of genres and adjustment of tempo in a continuous stream, positioning it as a tool for jamming with an AI. - The Lyria 3 integration in Gemini showcases a multi-tool AI workflow by automatically creating album art for the generated music using Google's built-in image generation model. - Competing models in the AI music space include Suno, which focuses on end-to-end song generation with vocals, and Stable Audio, which offers more granular, technical controls for detailed audio manipulation.
