Pull any song apart.
Cut clips, then split any song into vocals, drums, bass, guitar and piano. It all happens on your own computer: your music is never uploaded.
Three things, done well
Cut & join
Drag across a waveform to select seconds 5 to 7, delete them, and the rest joins up seamlessly. Undo anything.
Split into stems
A machine-learning model separates the song into its parts. Solo, mute and remix them.
Download anything
Each stem as WAV or MP3, all stems in one ZIP, or just the mix you're hearing (like a karaoke version).
Your device does the work
Pick a model
The first time, your browser downloads the model (60–91 MB) and keeps it. After that it loads instantly.
Drop a song
The file is read inside your browser. Nothing is sent to a server; there is no server doing the splitting at all.
Split, listen, download
Your graphics chip runs the model (about 3–6× faster than real time on a recent laptop). Then mix and export.
What you need
| Browser | Chrome or Edge 113+, or Safari 26+. These use your GPU (WebGPU) and are fastest. Other browsers still work, on the CPU, several times slower. |
|---|---|
| Computer | A laptop or desktop with 8 GB of memory or more. Phones aren't supported yet. |
| Storage | 4-stem model 91 MB · 6-stem model 60 MB · stored in your browser, removable any time |
| Mac app | Coming soon. Needs macOS 15+ on Apple silicon (M1 or newer). |
Good to know
Is it free?
Yes. The processing happens on your computer, so there's nothing to pay for.
Is my music uploaded anywhere?
No. Songs are opened and processed inside your browser tab. Only the model file is downloaded, once.
How good is the separation?
Vocals, drums and bass come out very clean. Guitar and piano are harder and may carry a little of other instruments. Busy, heavily produced songs are the hardest.
Can I use the stems in my own music?
Splitting songs for practice, study or fun is fine. Publishing stems taken from someone else's song still needs their permission.