ACE-Step

ACE-Step is an open-source music foundation model developed by ACE Studio and StepFun, and version 1.5 is widely seen as the strongest model you can run locally. It turns text descriptions into full songs with vocals in more than 50 languages, up to 10 minutes long, with style control through tags plus covers, repainting sections and vocal-to-instrumental conversion. It runs on consumer hardware, including NVIDIA and AMD GPUs and Apple Silicon Macs, and plugs into ComfyUI. Its license permits commercial use, which makes it popular with developers building their own music products. Hosted versions are available from third-party APIs for people who don't want to run it themselves.