By ShinobiTools Team · Last updated: July 2026
Upload a song and pull the vocal out as its own clean track — for sampling, transcription, mashups or just hearing the singer on their own. Free, no account, no watermark.
Split your track free →StemGrab runs Demucs, a hybrid-transformer source-separation model, on our own GPU, so there's no per-track fee and no watermark: upload a song and get seven stems back — vocals, drums, bass, guitar, piano, other, plus a clean instrumental — as 192 kbps MP3 files and a zip. No account, no e-mail, and your upload is deleted the moment it's processed.
Extracting the vocal doesn't cost you the rest of the song: the same upload also gives you the instrumental and each instrument on its own track, so you can rebuild or remix around the voice however you like.
An isolated vocal opens up things a full mix can't: dropping the acapella over a new beat, checking exactly what was sung, or lining lyrics up for captions.
If you've ever hunted forums and Discord servers for a decent acapella of a track, you know most of them either don't exist or come with the beat still bleeding through. Extracting the vocal yourself skips the digging entirely: upload the track you already own and you get the acapella out in under a minute, without a kick and snare ghosting under the voice on the drops.
Because the separation works on the original recording, the vocal comes out in the song's original key and tempo — nothing is re-pitched or time-stretched. That matters for mashups: if you want the acapella to sit over a different instrumental, do the key- and tempo-matching yourself in your DAW (Ableton, Serato, Rekordbox), where you can warp and pitch it deliberately rather than fighting a version that's already been altered. One caveat worth stating plainly: an isolated vocal is still the original master recording. Pulling it out is fine for private practice, bootlegs and study, but releasing or monetising a mashup means clearing the sample with whoever owns the song — the tool can't grant you that right.
Vocals, drums and bass come out remarkably clean on most modern, well-mixed tracks. Guitar and piano are newer, experimental stems and can be rougher, especially in dense mixes; lo-fi, live or heavily reverbed recordings are harder for any tool. Give the tricky ones a listen before you build on them.
Want just the backing instead? The instrumental maker covers removing the voice and keeping the band, and split a song into tracks gives you every stem at once.
The first time people solo a vocal stem, the reaction is often "it sounds thinner than the song". Nothing went wrong — you're hearing the vocal without everything that used to hold it up:
Fixing it is small work: a gentle high-pass, a touch of compression, and — if you're placing the vocal in a new track — add your own reverb rather than trying to remove what's there. Don't chase brightness with heavy EQ; you'll only bring the artefacts up with it.
Separation quality depends more on the file you start with than on any setting, and this is the part people skip:
What you're allowed to do with the result is a separate question from what's technically possible — the short version is that separating a track you own for practice or private use is a different matter from releasing it. That's covered properly in the guide to stems, sampling and copyright.
Yes — the vocals come out as a separate MP3, with no drums, bass or backing underneath on most modern mixes.
Very clean on well-mixed studio tracks. Dense, live or heavily reverbed recordings leave more bleed, which is normal for every separation tool.
Seven per song: vocals, drums, bass, guitar, piano, other, and a mixed instrumental (everything except the vocals), all as 192 kbps MP3 plus a zip of the lot.
Yes. The vocal is lifted straight out of the original recording, so its pitch and timing are untouched. If you want it over a different beat, warp and pitch it yourself in your DAW so you stay in control of the match.
The extracted vocal is still the original master recording. Extracting it for private practice, study or a bootleg is one thing; releasing or monetising a track built on it means clearing the sample with the rights holder. The tool separates audio — it can't grant you licensing rights.
No. Your upload is deleted the moment it's processed and the stems are wiped from the server within 45 minutes. Nothing is stored, shared or used to train anything — see privacy.