Free tool

Acapella extractor. Pull the centre vocal out of a stereo song, free, in your browser. No sign-up.

Updated

Files over 10:00 are trimmed to the first 10 minutes.

How to extract the vocals from a song

  1. Step 1: Drop in a stereo song

    An MP3, WAV, M4A or FLAC of a finished mix. It is opened on your device and never uploaded.

  2. Step 2: The centre is kept

    Lead vocals sit dead centre. The tool keeps, band by band, only what is the same in both channels, and drops everything below 120 Hz, where the bass and kick live.

  3. Step 3: Download the vocal

    Compare before and after, then save a WAV or a 192 kbps MP3. For a studio-clean a cappella, use AI stem separation instead.

At a glance

Acapella Extractor specifications
InputA stereo mix. WAV, MP3, M4A/AAC, FLAC, OGG and AIFF, or the sound track of an MP4 or MOV video. Decoding is the browser's own, so a format it can't read is refused, never guessed. Mono files are refused.
MethodCentre-channel extraction: per frequency band, the part both channels share; nothing below 120 Hz. Not AI separation
LengthThe first 10 minutes; a longer file is cut to that, not refused
File sizeUp to 100 MB for an audio file; a video up to 4 GB if its sound is under 20 minutes
OutputThe centre as a mono 16-bit WAV or a 192 kbps MP3, with before and after players to compare
Runs on your deviceYes. The file is read and processed in your browser tab and never uploaded
WatermarkNone
Sign-upNot needed
Daily limitNone: there is no usage counter or cap

How it finds the vocal

In almost every mix the lead vocal is panned dead centre, so it is identical in the left and right channels, while guitars, keys, backing vocals and the reverb are spread across the stereo field. The free Vocal Remover uses that to cancel the centre. This tool does the opposite and keeps it.

Keeping the centre takes more than adding the two channels together, which only turns the sides down by half. The song is split into 1025 narrow frequency bands, about every 12 milliseconds, and in each band the tool compares the two channels: where they match in level and timing, the band is kept; where they differ, it is dropped. Below 120 Hz everything is dropped, since the bass and kick sit there and a voice carries little.

Its limits, honestly

It keeps whatever is in the centre, not just the voice. The kick and snare, a centred bass's upper harmonics, a lead synth or a solo instrument panned centre all come through with the vocal. It loses what makes a vocal wide: stereo reverb and delay, double-tracking and harmonies panned left and right fade out, so the result is drier and narrower than the vocal in the mix.

On a sparse mix with a dry, centred vocal (a lot of pop, hip-hop and acoustic music) the result can be usable for a remix sketch or practice. On a dense mix it sounds like the vocal with the band's ghost behind it. A mono file is refused, because in mono everything is in the centre.

The clean way: AI stem separation

A separation model listens to the music rather than the panning, so it can pull the voice away from a centred snare or synth and keep the vocal's own reverb. In the SoundGen studio it splits a song into vocals, drums, bass and other instruments as separate tracks. A standard separation costs 6 credits, and new accounts start with 100 free credits.

Questions

Is it free?
Yes, with no sign-up and no daily limit. The processing runs on your computer.
Is this AI vocal isolation?
No. It is a signal-processing method that keeps what is identical in both stereo channels, which is usually the lead vocal. It is instant and free, but it can't tell a voice from a centred drum. AI stem separation in the SoundGen studio can.
Why can I still hear drums or music?
Anything mixed dead centre stays with the vocal: often the kick and snare, a lead synth or a solo. Only an AI separation model can split those apart.
Why does the vocal sound dry?
Stereo reverb and delay differ between the left and right channels, so they are dropped with the other off-centre sounds. Double-tracked and harmony vocals panned wide go too.
Why does it say my file is mono?
In a mono file both channels are the same, so everything counts as centre and there is nothing to separate by. AI stem separation works on mono files.
How do I get a clean a cappella?
Use AI stem separation in the SoundGen studio: vocals, drums, bass and the rest come out as separate tracks. A standard separation costs 6 credits; new accounts start with 100 free credits.
Is my audio uploaded?
No. The file is decoded and processed inside your browser tab and never leaves your device.
How long can the song be?
The first 10 minutes are processed; a longer file is cut to that rather than refused. Audio files can be up to 100 MB.

Want the vocal alone,
not the centre?

SoundGen's AI stem separation splits any song into vocals, drums, bass and the rest, straight onto tracks in a full studio.