AUD · Audio tools

Vocal Remover

Shareable link

Settings are written to the URL as you change them. Nothing differs from the defaults yet.

Remove the centre of a stereo mix

A lead vocal is usually panned dead centre, which means the same signal appears in the left channel and the right channel at equal level. Subtract one from the other and it disappears. Everything panned away from centre has a difference between the channels and survives.

That is the whole mechanism, and knowing it tells you exactly when this works.

Protect below and above

Kick, bass, and snare are also panned centre in most mixes. A naive subtraction takes them with the vocal, which is why the classic version of this trick leaves a track with no bottom and no backbeat.

The two protect controls set the band the cancellation runs inside. Below the first frequency, at 120 Hz by default, the source passes through untouched, so the bass and the kick stay. Above the second, at 12 kHz, cymbals and air stay. Only what sits between the two is cancelled.

Raise the low corner towards 250 Hz on a bass-heavy mix. Lower the high corner to 8 kHz when the vocal’s sibilance is what is left over.

Strength and mix

Strength sets how much of the banded centre is subtracted. At 100% the centre is fully removed and the two channels come out in opposite polarity to each other. Backing off to 70 or 80% often sounds better: the vocal drops far enough to sing over, and the mix keeps more of its body.

Mix blends the result against the original. It is the control to reach for when you want the vocal quieter rather than gone.

The mono fold

Removing the centre leaves the left channel holding the side signal and the right channel holding its inverse. Summed to mono, those cancel each other and the track drops towards silence. Phone speakers, smart speakers, and some club systems do exactly that.

Check the result in mono before you rely on it. If it collapses, lower the strength.

Where it fails

Reverb and delay on a vocal are usually stereo, so they stay after the dry vocal has gone. Doubled or panned vocals never cancel at all. Live recordings and anything mixed narrow have too little side content to separate.

For a cleaner instrumental than phase cancellation can give, a trained separation model is the only real answer, and it is not something that runs in a browser tab.

Isolate

The same maths inverted: keep the banded centre, drop the rest. In this mode the two frequency controls act as the band the vocal is taken from, so the bass below the low corner is filtered out of the result rather than protected. What comes through with the vocal is whatever else sits centred inside that band, usually the snare. Good enough to check a lyric or line up a timing reference, not good enough to release.

Frequently Asked Questions

It cancels whatever sits in the centre of a stereo image, which in most mixes is the lead vocal. It reads WAV, MP3, M4A, AAC, Ogg, Opus, FLAC, AIFF, and WebM audio, and exports in the format the source arrived in.

No. It is phase cancellation, which uses the difference between the left and right channels. There is no model here separating instruments, so the result is rougher than a trained separator and the trade is that it runs instantly on any length of file.

Cancellation needs two channels to compare. A mono file has no side signal, so there is nothing to subtract, and the file passes through untouched rather than turning into silence.

Kick and snare are usually panned centre alongside the vocal, so they cancel with it. The protect-below control is the fix for the kick and bass. There is no fix for a centred snare, which is the main reason the result sounds thin on some songs.

They set the band the cancellation runs in. Below the first frequency and above the second, the source passes through untouched, which keeps low end and cymbals in the mix even at full strength.

Subtracting the centre from both channels leaves the two sides in opposite polarity. That is the technique working as intended, but it means the result collapses towards silence if a player sums it to mono. Check on a phone speaker before delivering anything.

It keeps the banded centre and drops everything else, which gives a rough vocal-only track. Anything else panned centre comes with it, so treat the output as a reference rather than a clean stem.

Wide stereo mixes with a hard-centred vocal: most pop and rock from the 1970s onward. Narrow mixes, mono recordings, live records, and anything with heavy stereo reverb on the vocal work poorly.

Explore Our Tools

Browse all tools