HOW TO GET AN ACAPELLA FROM ANY SONG. just the voice, please

An acapella is a song’s vocal on its own: the lead singer, and usually the harmonies, with no instruments under them. Studios keep one for every song they record, but you rarely get to hear it. What you have is the finished mix, with the voice glued to the band.

An acapella extractor pulls them apart. The old trick was to cancel out whatever sat in the middle of the stereo mix. It sometimes worked on a karaoke track, but for an acapella it was hopeless: it removed the voice instead of keeping it, and took the bass and the kick drum along too. Modern extractors use an AI model trained to recognise each kind of sound, and split the song into separate tracks, called stems.

Chord Lens splits a song into six stems: vocals, drums, bass, guitar, piano, and Other for everything else. The vocal becomes one fader among six. Solo it and you have the acapella; mute it and you have the backing track. Because it works from the recording you’re listening to, a cover, a live take or a song that never had an official acapella splits the same way.

Singers use an acapella to hear exactly what the lead is doing with a phrase: where they breathe, where they slide into a note, how they shape the end of a line. Choir members and backing singers use it to pick out a harmony the band hides. Players use it to transcribe a melody without guessing which notes belong to the guitar.

You need a computer with Chrome or Edge and hardware acceleration on: the split runs in your browser, on your own machine, not on a server. The first split sets the splitting engine up once, a download of about 140 MB, and after that each song splits while you watch.

In the screenshot the demo song, Coming Home, is split and Vocals only is pressed. The S on the Vocals row is lit, so only the voice plays, and the drums, bass, piano, guitar and Other sit silent below it with their levels untouched.

On most songs the voice comes away cleanly: clear enough to hear every word, every breath and every harmony. It isn’t the studio’s own vocal track, though, and some songs show it more than others.

Reverb and echo. A vocal drenched in reverb leaves a little of its tail in the other stems, and a little of the band can ring through in the vocal. You’ll hear it most in the gaps between lines.

Backing voices. Harmonies and stacked backing vocals can land in the Other stem rather than Vocals, because they blend with pads and keyboards. That’s why soloing Other alongside Vocals brings them back. If the lead alone is what you want, leave Other out.

Crowds and choirs. A live audience singing along, or a full choir, is hard to pull apart from the lead. A studio version of the same song nearly always gives a cleaner acapella than a live one.

Old and distorted recordings. Mono records, heavy distortion and lo-fi production split less cleanly. You’ll still hear the melody, with more of the band bleeding through.

For learning a part, none of this matters much. A faint ghost of a guitar under the voice doesn’t stop you hearing where the melody goes. If it bothers you, try a different upload of the same song before anything else.

Loop one phrase at a time. A line of a melody is usually two to four bars. Loop it, listen three times without singing, then sing it with the record, then without. In the Studio the loop snaps to the beat, and you can save it with a name to come back to.

Slow the hard runs. Step the BPM down and the voice slows without dropping in pitch, from half the song’s tempo up to double. At 0.75× you can hear which way every note of a run goes. Our guide to slowing a song down without changing the pitch has the full routine.

Move it to your key. If the singer sits too high or too low for you, move the key up to 12 semitones either way. The vocal stem moves with it, so you can learn the melody where you’ll actually sing it. Small moves sound most natural. See the best key for your voice for how to pick.

Then sing over the band. Once you have the line, flip the mix: Karaoke mutes the vocals and Other, and you sing the part over the band with the words in time on the chart. Making a karaoke version covers that side.

Chord Lens is for practising and studying a song yourself: learning the melody, working out a harmony, checking the words, singing along. That kind of personal use is broadly considered fair use under US copyright law, with similar rules elsewhere.

Stem downloads, MP3 on Premium and WAV on Pro, are for that same personal practice and study, not for redistribution. Don’t share an acapella, sell it or put it in something you publish. If you want to release a remix, a mashup or a cover built on the original vocal, you need a licence from whoever owns the recording, such as a sync licence, or the platform’s own licensing where it offers it.

On Free, the acapella plays in Chord Lens and there is no download, which covers most of what a singer needs: the voice, looped, slowed and in your key, as often as you like.