AI Vocal Remover
Remove vocals from any song and keep a clean instrumental for karaoke — or pull the vocal out on its own as an acapella. Two tracks, about a minute, no install.
- Vocals + instrumental
- MP3 320kbps
- MP3 · WAV · OGG · FLAC up to 100MB
- Pro plan · ◆ 4 credits
Drop in the song
You get back two files: the vocal on its own, and the song without it.
Preview & Download
Once processing is complete, preview and download your separated stems.
Upload a file to get started
How to remove vocals from a song
An AI vocal remover works differently from an equaliser: a model trained on a large catalogue of music listens to the mix and rebuilds the voice as its own track. What is left over is the instrumental — so the vocal remover gives you both halves of the song, not just a muted copy.
Drop in the song
MP3, WAV, OGG or FLAC, up to 100MB. Higher-bitrate sources separate more cleanly than a 128kbps rip.
The model finds the voice
It works on timbre and harmonic structure, not on frequency cutoffs — which is why it keeps the bass and cymbals that a karaoke filter would eat.
Two files come back
The vocal on its own, and the same song without it. Play either in the browser or download both.
What people use a vocal remover for
Two tracks out of one file — the singing, and everything else.
Karaoke from any track
Strip the lead vocal and keep the full backing. Works on tracks that were never released as an instrumental.
Acapella for remixing
Pull the vocal out clean so you can drop it over a different beat, chop it, or pitch it.
Practice and rehearsal
Singers practise over the real backing track; players practise against the real vocal instead of a click.
Cover versions
Use the original instrumental as a reference bed while you record your own take on top.
AI vocal remover vs. karaoke filter vs. stem splitter
Three ways to get the singing out of a song. They are not the same, and the old one is why many people think vocal removers sound bad.
Old karaoke "voice remover" filters
They cancel whatever sits in the middle of the stereo mix. The lead vocal goes — and so do the kick drum, the bass and the snare, while reverb and backing vocals stay. That is the hollow, phasey sound most people remember.
AI vocal remover (this page)
A model trained on separated recordings listens for the voice itself and rebuilds it as its own track. The drums and bass stay in the instrumental, and you get the vocal as a clean acapella too.
Stem splitter
Same idea, more cuts: vocals, drums, bass and everything else as four tracks. Use it when you want to rebuild an arrangement, not just mute the singer.
Getting a clean instrumental: what affects the result
The model does most of the work, but the song you give it decides how clean the result sounds.
Source quality
A lossless file or a 320kbps MP3 separates cleanly. A low-bitrate rip has compression artefacts wrapped around the voice, and some of them stay behind.
Reverb and effects on the voice
A dry vocal comes out almost perfectly. Long reverb tails and heavy delay smear the voice into the rest of the mix, so you may hear a faint echo in quiet parts.
Backing vocals and harmonies
Stacked harmonies are voices too, so they come out with the lead. If you want to keep the choir and drop only the lead singer, no tool does that reliably yet.
How busy the mix is
Vocals over a full band separate well. A voice and one acoustic guitar playing the same notes is the hardest case, because they overlap in every way the model listens for.
Remove vocals from songs you made here
No download, no upload
Under any song you generated — on Create or in My Music — tap Remove vocals. The song opens here already loaded.
Make your own karaoke version
Generate a song with lyrics, then strip the vocal and sing it yourself — at a party, in class, or to record your own take.
Reuse the instrumental
Keep the backing track from a song you like as background music for a video, or as a bed to write new lyrics over.
Vocal remover FAQ
Is the vocal remover free?
It is part of the Pro plan (from $7.99 a month) and costs 4 credits per song. Both files — the instrumental and the acapella — are yours to download.
Is a vocal remover the same as a voice remover?
Yes, people use both names for the same thing: taking the singing out of a song. Older "voice remover" tools use a karaoke filter that also removes the bass and drums; this one uses an AI model that keeps them.
How long does it take to remove vocals from a song?
Usually about a minute for a normal-length song. Longer files take a little more.
Can I remove vocals from a YouTube video?
You need the audio as a file — MP3, WAV, OGG or FLAC. We do not download from YouTube or other sites; use a recording you have the rights to.
How is this different from the stem splitter?
Same engine, different output. The vocal remover gives you two files — the vocal and everything else — and it is faster and cheaper because it only has to make one cut. The stem splitter breaks the song into four separate tracks (vocals, drums, bass, other), which you want when you are rebuilding an arrangement rather than just muting the singer.
Will there be bleed left in the instrumental?
Usually a little, and how much depends on the mix. Heavily processed vocals, big reverb tails and backing harmonies are the hardest cases — you may hear a faint ghost in the quiet parts. Dry, forward vocals over a busy mix come out cleanest.
Does it work on live recordings?
It runs, but expect worse results. Crowd noise and room bleed sit in the same frequency range as the voice, so the model has less to separate. Studio recordings are what it was trained on.
What files can I upload?
MP3, WAV, OGG and FLAC up to 100MB. Higher-quality sources separate better — a 320kbps MP3 or a lossless file gives the model more to work with than a 128kbps rip.
Can I publish what comes out?
Removing the vocal does not change who owns the recording. For anything you release or monetise you still need the rights to the original. Personal practice, private karaoke and study are the usual safe uses.
Get the instrumental now
Drop the song in above. Two files come back in about a minute: the vocal on its own, and the track without it.