Google's Lyria 3: make 30-second audio tracks using text or images (in beta) (blog.google)

Google DeepMind has launched Lyria 3, a generative music model now available in beta within the Gemini app, allowing users to create custom music simply by providing a prompt or uploading a photo. This new feature significantly enhances creative expression by generating lyrics, musical styles, and tempos based on user input. From nostalgic afrobeat tracks to comical R&B jingles, Gemini can produce 30-second compositions with tailored cover art, aimed at making music creation accessible and enjoyable rather than perfectly polished. The Lyria 3 model improves upon its predecessors by enabling users to have more control over musical elements and generating more complex audio compositions without needing to provide their own lyrics. Additionally, all tracks are embedded with SynthID, a technological watermark that aids in identifying AI-generated content, thus reinforcing transparency in AI-generated music. This development not only empowers content creators on platforms like YouTube but also reflects Google’s commitment to responsible AI deployment within the music industry, balancing innovation with respect for copyright regulations and artistic integrity.