What should it sound like?
Pick a template or describe a sound.
Start from a template
Generating…
0:00
First run loads the model (up to a minute). Long songs take longer.
Library
Everything you generated — here and from Claude. Older files are cleaned up automatically.
Admin
Team accounts, GPU and model updates.
Team members
Server
Model updates
Every night the server checks GitHub and Hugging Face for new ACE-Step / Stable Audio 3 versions, builds them next to the current ones, runs a test generation with every model and only switches if all pass.
App deployments
Every change pushed to GitHub main is tested by CI, then deployed here automatically within ~5 minutes: pre-flight test → waits for running generations → update → health check → automatic rollback if anything fails.