Routes audio generation requests to the right backend: Suno or Udio for music beds and jingles, MiniMax or ElevenLabs for voiceover, AudioCraft for sound effects. The workflow is straightforward: reads project metadata to pick the model, assembles a prompt tuned to that provider's format, then calls a unified dispatcher that drops an MP3 into your project folder. The hard constraint is that TTS runs your script verbatim, so you proof it first. Works well when you need a 30 second coffee shop jingle or a product demo voiceover and want to stay inside the Claude context instead of opening three different web UIs.
npx skills add https://github.com/nexu-io/open-design --skill audio-jingle