Online Vocal Remover

mp3 → wav × 2

Cancel the centred vocal in a stereo song to make a karaoke instrumental, or isolate it instead. Centre-channel cancellation runs in your browser with the Web Audio API.

Settings
WorkspaceNothing is uploaded
Drop a song or audio file here
Supports MP3, WAV, FLAC, M4A and OGG

About this vocal remover

This tool decodes your song with the Web Audio API at 44.1 kHz, then averages the left and right channels to find the centre of the mix, the part most mixes place equally in both channels. A zero-phase high-pass filter is applied to that centre above 120 Hz, so the filtering does not shift the timing of what is left. The source file has to be stereo: a mono file has no left and right to compare, so the tool refuses it instead of guessing. This is plain arithmetic, not a machine learning model, and nothing beyond the page itself is downloaded to run it.

Separation mode picks which files come out: both tracks, the instrumental only, or the isolated vocal only. Vocal cut intensity only affects the instrumental track, from 20% to 100% of the filtered centre subtracted from each channel. The isolated vocal file is not scaled by this slider. It is always the full centre signal above 120 Hz, so it also carries anything else mixed to the centre, such as a centred snare or solo. Content below 120 Hz is left untouched on purpose, so the kick drum and bass stay in the instrumental track. Both output files are written as 16-bit WAV at 44.1 kHz, regardless of the sample rate of the file you started with.

How to use it

01

Add one song file

Drop a stereo MP3, WAV, FLAC, M4A or OGG file on the workspace, or click Select audio file. Only one file is processed at a time.

02

Set separation mode and intensity

Pick both tracks, instrumental only or vocal only from Separation mode, and set Vocal cut intensity from 20% to 100%. Intensity only changes the instrumental track.

03

Click Separate vocals and instrumental

Your browser decodes the file, cancels the centred signal above 120 Hz and writes the result as WAV.

04

Preview and download

Play each track with the built-in audio player, then download the instrumental, the vocal, or both, depending on the mode you picked.

What it does

  • Centre-channel cancellation runs with the Web Audio API alone; no library is downloaded to do it.
  • A zero-phase high-pass filter keeps content below 120 Hz out of the cancellation, so the kick drum and bass stay in the instrumental track.
  • Vocal cut intensity adjusts only the instrumental track, from 20% to 100% of the centre subtracted.
  • The isolated vocal file is always the full centre signal above 120 Hz, not reduced by the intensity slider.
  • Both output files are written as 16-bit WAV at 44.1 kHz, whatever the sample rate of the source file was.
  • Separation mode produces both files, the instrumental only, or the isolated vocal only.

Specifications

Specification Details
Accepts Stereo MP3, WAV, FLAC, M4A or OGG that your browser can decode
Requires A stereo file. A mono file is rejected with an error, since there are no left and right channels to compare
Low cut 120 Hz. Content below this is left out of the cancellation so the kick drum and bass survive
Vocal cut intensity 20% to 100%, applied to the instrumental track only
Output format 16-bit PCM WAV at 44.1 kHz for both tracks, regardless of the source sample rate
Output files The source file name with _instrumental.wav or _vocals.wav added, depending on Separation mode
File limit One file at a time. There is no fixed size limit, but a long song uses more of your device's memory
Libraries None. Only the browser's built-in Web Audio API
Good to know

What the vocal remover does not do

This is centre-channel cancellation, not a source-separation model, and it has real limits worth knowing before you rely on it.

Needs a centred, dry vocal

It works best on older or simply mixed recordings. A modern mix that widens the lead vocal, doubles it, or adds heavy reverb will not cancel cleanly, leaving a quieter ghost of the voice rather than silence.

Cancels more than the voice

The technique has no idea what a voice is. Anything else placed dead centre, such as the snare or a centred lead instrument, is subtracted along with it. Only content below 120 Hz is protected.

Not a clean acapella

The isolated vocal file is the centre of the mix above 120 Hz, which includes the voice plus anything else centred. It is useful for following a lyric or a melody, not as a clean vocal stem.

Weaker in mono playback

A phone speaker or a mono PA adds the left and right channels back together, which brings the cancelled content back and weakens the instrumental track, especially at a high intensity setting.

Not a source-separation model

A trained source-separation model handles modern, reverb-heavy mixes far better and can produce real instrument stems. Those models need much heavier processing than a browser tab, and most web versions of them upload your audio to a server to run it.

Related tools

MP3 to WAV Decode to uncompressed PCM for editing. WAV to MP3 Shrink a session export for sending. MP4 to MP3 Extract the audio track as a 192 kbps MP3. Audio converter Batch convert between the four common audio formats.
All audio and video tools → How centre-channel vocal removal actually works →

Frequently asked questions

Yes. It cancels the signal shared between the left and right channels, so a mono file, which has no sides to compare, is rejected with an error instead of processing silently.

No. Vocal cut intensity only changes how much of the centre is subtracted from the instrumental track. The isolated vocal file is always the full centre signal above 120 Hz, wherever the slider is set.

No. It is centre-channel cancellation, arithmetic that runs with the Web Audio API. There is no model, no training data and nothing to download beyond the page.

Often only partly. It does best on older or simply mixed recordings with a centred, dry vocal. A mix that widens or doubles the lead vocal, or adds a lot of reverb, leaves a ghost of the voice instead of silence.

Because everything centred above 120 Hz is removed, not only the voice. A centred snare or lead instrument goes with it. Lowering the intensity leaves more of the mix in, at the cost of leaving more vocal too.

No. Decoding, the centre-channel arithmetic and the WAV encoding all run in this tab with the Web Audio API. No library is downloaded to do it.

Your files stay on your device

Your browser decodes the song and runs the centre-channel arithmetic inside this tab. Nothing is uploaded, and no library is downloaded either: separation uses only the browser's built-in Web Audio API.