Vocal Remover & Stem Separator

Split a song into vocals, drums, bass and other instruments with Demucs, one of the best open-source separation models, running in your own browser. Download an instrumental, an acapella or all four stems, and your music is never uploaded.

Your files stay on your device: they are processed in this browser tab’s memory and never uploaded.

Output
Quality
Smoother uses more overlap between segments, and takes about twice as long.

The Demucs model is about 170 MB and is downloaded once. Keep this tab open while it works.

How it works

The song is cut into overlapping segments of about eight seconds, and the Hybrid Transformer Demucs model separates each one into four stems. The segments are then blended back together and you can listen to each stem before downloading it.

Speed and hardware

The model is about 170 MB and downloads once. With a GPU in a recent Chrome or Edge, a song takes about a minute or two. Without a GPU it still works, but a full song can take a long while, so keep the tab open.

FAQ

Which stems do I get?

Vocals, drums, bass and other, plus a ready-mixed instrumental without the vocals.

What format are the stems in?

WAV at 44.1 kHz stereo, which is ready for any editor or DAW.

Is my song uploaded?

No. The separation runs on your device.

Can I use the separated stems in my own release?

Only if you have the rights to the original song. Separating a track does not change who owns it.

What can you do with a vocal remover?

Editors get a clean instrumental that sits under a voice-over; singers make karaoke and practice tracks; DJs and producers pull acapellas and drum loops for remixes; and musicians isolate a bass line or a drum part to learn it by ear.