Sing over a clean backing track
Strip the lead vocal out of a song and you are left with a full instrumental to sing over. No muffled center channel or ghost of the original singer bleeding through — just the band, ready for your own take on top.
Take the singing out of a track for a clean instrumental, or isolate the acapella to build on. One split gives you both — full source separation in the browser, with nothing to install.
Free tier · 30 credits / month · Lossless WAV export.
Drop any track onto the canvas — a finished mix, a demo, a downloaded file. Nothing to install and nothing to configure first; the song lands as a card and is ready to work on the moment it finishes uploading.
Run the split. A trained model listens to the mix and pulls the singing apart from everything behind it, so the vocal lifts out cleanly while the band, beat, and bed stay together as one instrumental.
Keep the clean instrumental for a backing track, the isolated acapella for a cover or remix, or both. Every part exports as lossless 24-bit WAV — no re-encoding between the split and the file you save.
Strip the lead vocal out of a song and you are left with a full instrumental to sing over. No muffled center channel or ghost of the original singer bleeding through — just the band, ready for your own take on top.
Pull the isolated vocal out of a track and you have an acapella to build on. Drop it over a new beat, chop it into a remix, or feed it straight into voice conversion to put the part in a different singer.
Solo the vocal to catch every word and inflection, or mute it to play along with the band. Hearing one element on its own makes a melody, a lyric, or a chord change far easier to transcribe by ear.
Grab a clean vocal phrase with none of the instrumental glued to it. An isolated sample sits in a new arrangement without dragging the original drums and bass along, so it pitches and chops without mud.
Use the vocal-free instrumental as a music bed so a voiceover or dialogue reads clearly. Without a competing lead vocal fighting your narration, the score sits underneath instead of stepping on the words.
Separate the vocal from a rough mix when you no longer have the session files. With the part isolated you can re-balance, re-effect, and re-place it over the instrumental instead of fighting a baked-in blend.
A trained model actually identifies the vocal and lifts it out, rather than inverting a channel and hoping the singing cancels. That is why the instrumental comes out full and present instead of hollow, thin, or smeared with artifacts.
One split hands you the instrumental and the acapella, not just whichever side you asked for. Keep the backing track and the isolated vocal from the same run — there is no second pass to extract the part you skipped.
When two stems are not enough, run a full six-stem split — vocals, drums, bass, guitar, piano, and other — and pull each instrument apart on its own.
See full stem separation →The instrumental, the acapella, and any individual stem all download as 24-bit WAV. Nothing is bounced to a lossy format between the split and your drive, so the parts hold up when you mix or master them later.
The acapella you just isolated drops straight into voice conversion and mastering on the same canvas. Split, swap the singer, and master without shuttling files between three separate apps and stitching them back together.
You pay for what you separate, not per login. The free plan ships with 30 credits a month and no credit card, so you can split a song into its instrumental and acapella before you decide on anything else.
Upload the track to the canvas and run the split. A trained model separates the singing from everything else and gives you a clean instrumental with the lead vocal pulled out. You can keep the isolated vocal from the same run too, then download whichever parts you need as lossless WAV.
No. The hollow, watery sound comes from the old channel-inversion approach, which guts the center of the mix. This uses real source separation that recognizes the vocal and removes only that, so the drums, bass, and everything else stay full and present. The instrumental sounds like a mix, not a karaoke cassette.
Yes. A single split produces both the instrumental and the isolated vocal, so you do not have to choose one and lose the other. Download the acapella on its own to build a cover or remix, or feed it directly into voice conversion on the same canvas to put the part in a different voice.
Upload a standard audio file and the parts come back as lossless 24-bit WAV. Everything runs in the browser, so there is nothing to install to handle a full-length song. If you need more than a vocal and an instrumental, the same upload can be run through a full six-stem split instead.
Yes, you can start free. The free plan includes 30 credits a month with no credit card required, which is enough to separate a song into its instrumental and acapella and export both. Billing is by credits rather than per seat, and commercial use is available on paid plans.
The classic trick subtracts one stereo channel from the other to cancel anything panned dead center — usually the lead vocal, but also the kick, snare, and bass, which all collapse with it. That is why the result sounds hollow and only works on some songs. Model-based separation listens to the music instead, identifies the vocal specifically, and lifts it out while leaving the rest of the mix intact.
30 free credits. No card. Pull the instrumental and the acapella, then convert and master — all on one canvas.