Full songs with vocals
Four minutes and past it, with lyrics you write or generate.
01 — Open source · AI music/Live
ACE-Step UI
ACE-Step 1.5 put good music weights in the open. What it did not have was software a musician would keep open for an afternoon. So I wrote that.
- Stars
- 4,748
- Active users
- 15k
- Social reach
- 1M+
- Licence
- MIT
Role
Design, frontend, backend, releases
Year
2026
Stack
- React
- TypeScript
- Express
- SQLite
- ACE-Step 1.5
A local Suno, and the interface it was missing
What it is
A full generation studio for ACE-Step 1.5, running entirely on your own machine. Songs up to four minutes with vocals and lyrics, instrumentals, style reference from any audio file, stem separation through Demucs, and a library you can actually search.
The repo description says it plainly: the ultimate open source Suno alternative. Free, local, unlimited. Nobody meters your downloads.
Why it caught on
The weights had been public for a while and most people still could not use them. Installing meant a Python environment, a Gradio tab, and a folder of untitled wav files. I gave it the shape people already know — a Spotify-style library, a player, playlists, live queue position, dark and light mode — and made install a single click through Pinokio.
It is now the interface a lot of the open-source music crowd runs by default.
The audience
Fifteen thousand people use it actively. The posts and clips around the releases have cleared a million in reach, which is where most of those installs came from — no ads, no launch budget, just the work shown in public while it was being built.
Four thousand seven hundred stars and seven hundred and thirty-three forks sit on top of that.
How you run it
Three ways in. Pinokio installs it in one click, which is what I point most people to. There is a Windows portable package, about 5GB, for anyone who does not want to touch a terminal. Or clone the repo and run the setup script.
It needs Node 18+, Python 3.10+ and FFmpeg. An NVIDIA card with 4GB of VRAM will generate; 12GB opens up the LLM-assisted features.
What is in it
Style reference
Point it at any audio file and generate in that direction.
Stems
Vocals, drums, bass and other, separated with Demucs.
Music videos
Pexels backgrounds cut against the track.
