Vocal remover

How the vocal remover works

Older vocal removers used a trick: most lead vocals are mixed to the centre, so subtracting the left channel from the right cancels them. It also cancels the bass, kick and snare, and it does nothing for vocals with stereo reverb. This tool uses an AI model instead. Demucs, an open-source model trained on thousands of multitrack recordings, listens to the whole song and rebuilds each part as its own track: vocals, drums, bass, guitar, piano and everything else.

The model runs inside your browser with WebAssembly, on your own computer or phone. Nothing is uploaded, there is no account, and there is no limit on how many songs you separate. The first run downloads the model (about 140 MB) and your browser keeps it, so later songs start straight away.

Karaoke, acapella or something in between

After separating, the quick buttons cover the common jobs, and the mixer below them lets you set each part by hand:

If you need more than vocals and music, for example the drums on their own or the song without its guitar, the stem splitter page covers the other instruments in detail.

Removing only the lead singer

Many karaoke singers prefer to keep the backing vocals: the harmonies fill out the track and help you stay in tune. Under Split further, Lead / backing vocals runs a second model (UVR’s MDX-Net Karaoke) on the separated vocals and splits them into the main singer and the backing voices. Mute Lead and keep Backing for a fuller karaoke track. This step takes longer than the main split, about three to four times the song’s length on a typical computer, so it’s worth trimming the song first if you only need part of it.

Make the karaoke track fit your voice

The same editor that removes the vocals can also move the song into your key. Open Speed & pitch, choose Set separately and drag the pitch slider: each step is one semitone, and the tempo doesn’t change. Two or three semitones down is often enough to make a high chorus comfortable. If you are learning the song, you can also slow it down while keeping the key; the slow down a song page explains the settings. Drag the handles on the waveform to trim an intro or loop a hard section.

Getting the cleanest result

• Start from the best file you have. A WAV, FLAC or high-bitrate MP3 separates more cleanly than a heavily compressed copy.
• Lower instead of muting. If a faint echo of the voice remains, turning the Vocals track down to a low level can sound more natural than muting it outright, especially on songs with lots of reverb.
• Mask leftovers with reverb. With the vocals muted, a little room reverb on the music hides small artifacts. Switch the vocal track’s Effects off if you keep it in the mix, so only the music is affected.
• Pick fewer instruments on a phone. Before separating, you can choose which parts to pull out. Picking only vocals uses less memory on long songs, though it takes the same time.

What people use vocal removal for

Karaoke and singing practice

The obvious one: an instrumental of the exact recording you know, rather than a karaoke cover that sounds slightly off. Singers also use a low guide vocal while learning a song, then mute it once they know the words. Loop the chorus by dragging the handles on the waveform and turning on Loop.

Choirs, worship teams and school concerts

Rehearsal tracks without the lead vocal let a soloist or a whole choir sing along with the original arrangement. Moving the key down a tone or two often makes a pop song singable for younger voices, and the tempo stays the same.

Acapellas for remixes and edits

Producers pull the vocal out to build a new track under it. Export it as 24-bit WAV and the timing stays sample-accurate, so it lines up with the original tempo in your DAW. The page shows an estimate of the song’s BPM to help you set the project tempo.

Transcribing lyrics and harmonies

A solo vocal track makes mumbled lyrics and stacked harmonies much easier to hear. Split Lead / backing vocals to listen to the harmonies on their own, and slow them down without changing the pitch if a line goes by too fast.

Formats in and out

The vocal remover opens MP3, WAV, FLAC, M4A/AAC, OGG, Opus, AIFF and WMA files, plus the soundtrack of MP4, MOV and MKV videos. Downloads can be MP3 (128–320 kbps), WAV (16-bit, 24-bit or 32-bit float), 24-bit FLAC or M4A, and only the tracks you left on are included. The title, artist and cover art from the original file are kept. Choose Video to make a 1080p MP4 for sharing; videos carry a small slowed-reverb.com credit in one corner, while audio files have no watermark.

Want to go further than karaoke? Separated tracks work with every effect on the site, so you can make a slowed + reverb edit of just the instrumental, or turn the piano part into a falling notes video.