On September 4, 2026, Google began providing the music generation model Lyria 3.5 via the Gemini app and Gemini API. The company stated that this model can produce high-fidelity music with expressive vocals and rich arrangements. In the Gemini app, users can select genres, choose between vocal or instrumental versions, and specify the length of the track. Templates for purposes such as BGM or birthday songs are also available.
For developers using the Gemini API, detailed control is possible, including structural instructions via lyric section tags and the specification of instrument entries/exits and fade-outs using timestamps. In addition to text, up to 10 images can be used as input. The system is designed to output the lyrics and song structure as text along with the generated track. The usage fee is $0.08 per song, and no free tier has been established.
Source: Google、音楽生成AI「Lyria 3.5」をGeminiアプリとAPIで提供開始 - innovaTopia (Google News: Gemini)