Vocal Remover

Split a song into the music and the vocals — for karaoke, practice or a remix. Runs on your device; the song is never uploaded

Drop a song here

MP3, M4A, WAV, FLAC, OGG or a video — processed on your device, never uploaded

Two files from one song
The instrumental (a karaoke or backing track) and the vocals on their own.
A trained separation model
SCNet, an open-source neural network, not the old trick of cancelling the center channel.
Songs and videos
MP3, M4A, WAV, FLAC, OGG, or the soundtrack of a video. Save as MP3 or WAV.
100% Private
The song is separated in your browser and never uploaded.

Remove vocals from a song, free and without uploading it

Vocal Remover splits a song into two files: the instrumental, with the singing taken out, and the vocals on their own. It runs SCNet, an open-source music separation model, in your browser, so the song stays on your device. There is no account, no queue and no limit on the number of songs.

How to make an instrumental or karaoke track

  1. Drop a song on the box above: MP3, M4A, WAV, FLAC, OGG, or a video.
  2. Click “Remove vocals”. The first time, the model downloads once (13 MB).
  3. Listen to both results and download the instrumental, the vocals, or both, as MP3 or WAV.

What it is good for

How good is it?

On studio recordings SCNet removes most of the singing, usually leaving faint traces on loud notes, reverb tails or backing vocals. Live recordings, heavy distortion and very old mono songs separate less cleanly. The vocals file can keep a little of the drums or guitar. The model was trained on the MUSDB18 music collection; on its test songs it separates vocals better than the widely used Demucs.

How long does it take?

Separation is heavy work for a browser: on a computer it takes a few times longer than the song itself, on a phone longer still, and it needs free memory, so a computer is the better choice for long songs. The page shows which part it is working on and about how much time is left.

Privacy

The song is decoded and separated in your browser with ONNX Runtime Web. The page downloads the model file (13 MB) and the runtime once; your audio is never sent anywhere. AudioTools.dev has no server that could receive it.

Need something else? Cut out a part with the audio trimmer, make it louder with the volume booster, or change the format with the audio converter.

Frequently asked questions

How do I remove vocals from a song for free?
Drop the song here and click Remove vocals. When it finishes, download the instrumental as MP3 or WAV. It is free, with no account and no limit.
Is my song uploaded to a server?
No. The separation model runs in your browser; only the model file is downloaded, once. The song never leaves your device.
Can I get the vocals only (acapella)?
Yes. You get two files: the instrumental and the vocals. Download either or both.
Does it work on a phone?
Yes, in current Chrome and Safari, but it is several times slower than on a computer and needs free memory. For long songs, use a computer.
Why can I still hear some of the voice?
A separation model estimates the voice; it cannot know it exactly. Reverb, backing vocals and loud notes can leave faint traces, especially in live or heavily processed recordings.