ACE-Step
Open-weight music models you run yourselfFree option
About ACE-Step
Open music generation model under Apache 2.0, with weights on Hugging Face and ModelScope. The README says v1.5 came out on 2026-01-28 and calls it the latest model. It lists 19 languages for text to music, plus lyric editing, audio repainting and instrumental generation. It needs 8 GB of VRAM at minimum with optimizations on, and the project recommends an A100, RTX 4090 or RTX 3090. The project asks users to verify originality and disclose AI involvement.
Everything here is what the tool's own site states on the date shown. It is not a test result. Spotted a mistake? Tell us.
More in Open-weight music models you run yourself
MusicGen and AudioCraftFree option