Summary
Four RunComfy endpoints for StepFun's ACE Step open-weights music model: text-to-audio generation, a 1.5 variant with 50+ language vocals, audio inpaint to regenerate time ranges inside existing tracks, and outpaint to extend before or after. You drive composition with tags (genre, mood, instruments, BPM) and optional structured lyrics with section markers. Pricing is $0.0002 to $0.0003 per second of output, which works out to about 27× cheaper than ElevenLabs Music for comparable generation. The skill is pitched as a draft-quality workhorse for high-volume iteration, game loops, and background music where cost matters more than polish. Inpaint and outpaint are rare capabilities in the music-gen space and genuinely useful if you need surgical edits without re-rendering a whole track.
Install
npx skills add https://github.com/doany-ai/skills --skill ace-step
