Make Music is a skill that enables an AI agent to generate audio tracks from text descriptions, using Google’s Lyria 3 models accessed through the Gemini API. It supports two model variants: lyria-3-clip-preview for short loops and background clips up to 30 seconds, and lyria-3-pro-preview for full song structures running 2 to 3 minutes. Both instrumental and vocal track types are supported.
The skill handles the API handshake, directory management, and file serialization internally, delivering a ready-to-play MP3 without requiring the developer to orchestrate those steps manually. Output files are 48kHz stereo and carry a SynthID watermark.
Developers can control generation parameters including BPM, musical key, specific instruments such as 808 drums or Rhodes piano, and mood descriptors. For structured compositions, timestamped prompts let an agent define transitions between sections like intros, verses, and choruses.
The skill uses the free tier of the Google GenAI SDK, so there are no per-track costs at that usage level. It is the right choice when an agent needs to produce actual audio files rather than symbolic representations like chord charts or lyrics. It is not suited to tasks that require only lyric writing, music notation, or audio editing of existing files, nor to workflows that cannot depend on the Gemini API and the Google GenAI SDK.