Rehearse your part with the band, minus the singer
Learning a melody on sax, getting ready for a lesson, drilling a choir part? RemoveVocal takes the voice out of a mixed song so you can play or sing along with the band that recorded it.
Split free: songs up to 20 min, played in full · Fine on a tablet at the music stand · MP3 results
- Band kept, singer out
- Original key and tempo
- Phone, tablet or laptop
- Up to 320 kbps MP3
Any song you're learning works: MP3, WAV, FLAC, M4A, or a video such as MP4 or MOV.
From the record to a rehearsal track
Pick the song you're learning
Upload the song your lesson, choir or band is working on. A music file or a video file you own both work.
The singer steps aside
The model separates the voice from the band and keeps both halves. You can watch progress while it works.
Play it through, then save
The player runs the whole song, so hop from original to vocals to instrumental as it goes. Save the tracks as MP3 and take them to practice.
Made for the practice room
It rebuilds, it doesn't subtract
Old tricks cancel the middle of the stereo mix and take the bass with it. The model rebuilds the band without the voice, so the low end stays where it was.
The singer as a reference
Taking the singer out also gives you the singer. Keep the isolated vocal to check how a melody or a lyric goes, then practice over the band alone.
Same key, same tempo
Nothing is transposed or stretched, so what you learn over the practice track matches the record when the band plays it live.
On the device on your stand
Windows, macOS, Linux, Android or iOS: if it runs a modern browser, it runs RemoveVocal. No install before rehearsal.
Practicing with the record, without the voice
The best way to learn a part is often to play it with the real band. A practice track gives you the recording you already know, minus the person singing, so your instrument or your voice takes the lead line. Making one is harder than it sounds: nothing in a mixed file marks which sounds are voice, and a belted chorus can share its frequencies with a distorted guitar.
RemoveVocal runs a separation model trained by InverseAI. It listens to the full mix, estimates which parts of the sound are a human voice and which are instruments, and writes two new recordings. Because it judges by the character of the sound rather than by stereo position, it copes with things that defeat simpler tools: stereo reverb on the vocal, doubled lines, or a singer panned slightly off center.
Backing vocals and harmonies count as voice, so they leave with the lead. If your arrangement needs the backing singers to stay in the band track, expect them to be removed too.
Lessons, auditions and rehearsals
Most people who need a practice track need it for something happening soon: band rehearsal on Thursday, an audition next week, a choir run-through on Sunday. They don't want to learn audio editing first. Upload, wait a few minutes or less for a typical song, listen, save.
Instrumentalists use it differently. A saxophone player mutes the singer to play the melody herself; a teacher strips a song so students can sing it in a lesson. For the sing-along side of this, the karaoke maker page has the details.
If what you need is separate drum, bass or guitar tracks, RemoveVocal is the wrong tool: it produces exactly two tracks. Our explainer on AI stem separation covers how multi-stem tools differ.
Where a practice track can fall short
Every separation leaves something behind now and then. You might catch a ghost of the voice on a long held note, inside a big reverb tail, or where the singer is mixed far louder than the band. Live bootlegs and old mono recordings are tougher than modern studio mixes.
A low-bitrate file already carries compression artifacts, and separating it tends to make them easier to hear. Use the original purchase, a CD rip or a lossless file when you can. Then listen on headphones in the player; most leftover traces disappear once you sing or play on top.
Take the result somewhere
Backing tracks for covers
What to check in an instrumental before you hit record.
Remix acapellas
Use the vocal half for remixes and transcription.
Already have an MP3?
Notes on bitrate and why the source copy matters.
Cleaner separations
How reverb, mixing choices and file quality change the result.
Learn it by ear first
Work out the melody and the chords before you rehearse them.
Three ways to get the singer out
A fair look at the methods, including where the older ones still make sense.
| AI separation (RemoveVocal) | Phase-cancellation trick | EQ in an editor | |
|---|---|---|---|
| Typical songs it handles | Most stereo mixes | Songs with a dead-center vocal | Few, and only partly |
| What happens to bass and kick | Kept | Often thinned out | Cut along with the voice |
| Do you also get a vocal track | Yes | No | No |
| Reverb and doubled vocals | Mostly handled | Left behind | Left behind |
| What you need | A browser | An audio editor | An editor and practice |
Phase cancellation flips one stereo channel and adds it to the other, so anything panned to the center fades, including bass and kick. EQ can only lower frequency ranges, and the voice shares those ranges with guitars and keys. Both are still handy for quick experiments on simple mixes.
Practice-room questions
Other rooms in the studio
Ask the singer to sit this one out
Upload the song and hear the band on its own, with the vocal saved separately if you want it.