Skip to main content
Edit Videos

Free · private · in your browser

Remove vocals from a video

Compare the original sound, choose the reduction strength, then export.

How it works
  1. ImportChoose source
  2. EditSet the change
  3. PreviewReview result
  4. ExportCreate file
InputMP4, WebM, MOV and M4V videoOutputMP4 or WebM video

Most music is mixed with the lead vocal in the centre of the stereo image and the instruments spread around it. That structure can be exploited: subtracting one channel from the other cancels whatever is identical in both, which is largely the voice. This tool analyses the mix, previews the reduction at different strengths, and exports the result.

Centre-channel cancellation and what it costs

The technique removes everything centred, not just the singer. Bass and kick drum are almost always centred too, and often the snare, so an aggressive setting can hollow out the rhythm section along with the voice. That is why strength is adjustable rather than fixed: a moderate setting reduces the vocal enough to sing over while leaving the backing recognisable, and the strongest setting removes more voice at real cost to the rest. Listen to the whole track rather than one chorus, since density varies across a song.

Which recordings work and which do not

Results depend almost entirely on the mix. Conventional stereo productions with a firmly centred lead vocal do well. Live recordings do poorly, because reverb spreads the voice across both channels and the reflections are not centred. Mono recordings cannot work at all — with identical channels there is no difference to exploit, so cancellation removes everything or nothing. Heavily processed vocals with wide stereo effects also resist the technique, since the voice is no longer purely centred. Expect reduction rather than removal, and judge by whether what remains is usable.

Questions, answered

Can this remove vocals completely?

Rarely. It reduces centred content substantially, which usually includes most of the lead vocal, but reverb and any stereo processing on the voice survive. Treat it as reduction, not removal.

Why did the bass disappear too?

Bass and kick are usually centred in a mix, so they cancel alongside the voice. Reduce the strength to keep more of the rhythm section at the cost of more vocal remaining.

Why does it not work on my recording?

Most likely the audio is mono, or the vocal is not centred — live recordings and heavily processed vocals both spread the voice across the stereo field, which defeats the technique.

Is this the same as AI stem separation?

No. This uses the stereo structure of the mix rather than a trained model. It is fast and runs locally, but a good separation model can achieve cleaner results on difficult material.

Can I use the result publicly?

Removing vocals does not change who owns the recording. Copyright still applies to the backing track, so check the rights before publishing.