Skip to content
All tools

Remove the vocal from a song

Cancel out anything sitting in the centre of a stereo mix — which is where the lead vocal almost always is — and keep the rest as a backing track. Works on the mix you have, on your own device.

Keep the bass and kick Low frequencies are taken from the original instead of being cancelled, so the backing track doesn’t go thin.

Ignored by the lossless formats.

Advanced 1
200 Hz

Bytes of your files sent anywhere: 0 B

How to use it

  1. 01

    Add your files

    Add a stereo song. Mono files can’t be processed this way, and the tool will say so.

  2. 02

    Check the settings

    Leave keep the bass on unless the result sounds better without it. It takes the low end from the original so the backing track doesn’t go thin.

  3. 03

    Save the result

    Press remove the vocal, then listen before you save — how well it works depends on the mix.

Why it doesn’t upload

This is arithmetic on the two channels, not a model that has to run on a server. Your browser subtracts one channel from the other, so the song never leaves your device.

How to check for yourself

Worth knowing

  • This is centre cancellation, not AI separation. It removes whatever is mixed dead centre. On most songs that is the lead vocal, but it also takes out anything else sitting in the middle.
  • It works best on older and simply-mixed recordings. Modern pop often widens the vocal or drenches it in stereo reverb, and anything not perfectly centred survives — usually as a ghost of the vocal.
  • The result is mono. That is unavoidable: the backing track is built from the difference between the two channels, which is a single signal.
  • Mono files can’t be processed at all. Subtracting a channel from itself leaves silence, so the tool refuses rather than handing you an empty file.
  • It cannot give you the vocal on its own. Isolating the centre needs a trained model, which this tool does not use.
  • Uses the FFmpeg engine, a one-time 31 MB download.

Questions

How does removing vocals without AI work?

A lead vocal is normally mixed equally into the left and right channels. Subtracting one channel from the other cancels anything identical in both, so the vocal disappears and sounds panned to one side or the other remain.

Why does the backing track sound thin?

Bass and kick drums are usually centred too, so they cancel along with the vocal. Keep the bass takes everything below about 200 Hz from the untouched mix instead, which puts the low end back. Raise the cutoff if the result is still thin.

Why can I still hear the singer?

The vocal was not perfectly centred — stereo widening, double-tracking or reverb spread across the mix all leave part of it in the difference signal. Nothing in the settings will fix that; it is a limit of the method, not of the tool.

Can I get just the vocal instead?

Not with this. Cancelling the centre can remove what is in the middle but cannot extract it on its own — that needs a trained separation model. This tool makes backing tracks.

Is the result mono or stereo?

Mono. The backing track is the difference between the two channels, which is one signal by definition. You can save it as any format you like, but the content is mono.

Does it work on a live recording?

Often better than on a studio track, since a live mix tends to put fewer things dead centre. As always, listen to the result before you rely on it.

Other tools