Free · In your browser · Nothing uploaded
Acapella extractor, made for sampling.
Drop a song in and get two files back: the acapella and the instrumental. Then do what you came for — chop the vocal into a phrase, a hook, a single syllable, and play it on TAIL's 16 pads. The separation runs on your own device; the song never leaves this tab.
How it works
How to extract an acapella from a song
Drop the song in
MP3, WAV, M4A, FLAC, OGG — or a video, the audio is taken from it. Up to ten minutes per song; a longer file is offered a split into parts.
Extract the acapella
The first time, the separation model downloads (78 MB, then cached). A four-minute song takes about a minute on a normal laptop.
Chop it or play it
Download the acapella and the instrumental, chop the vocal on the waveform, or load both onto TAIL's pads in one click.
What producers use it for
From a finished song to vocal chops
Vocal chops
The reason most people want an acapella. Chop a line into syllables, put each one on a pad, and play the vocal as an instrument instead of a loop.
Remixes and bootlegs
The acapella over your own beat. Match its tempo and key first with the BPM changer and key changer, then build around it.
Flipping the instrumental
Sometimes the sample is under the voice. The instrumental comes out as its own file, ready to chop like any loop.
Studying a vocal
Hear how a hook is stacked, where the breaths are, what the ad-libs do — without the mix in the way.
Acapella → pads
Every other extractor stops at the download
An acapella on its own is a file in a folder. Here it is the first step of a sampler: press Chop the acapella and it opens in the sample chopper, ready for chop markers. Or send the acapella and the instrumental straight onto TAIL's pads, and play them against each other in the browser sampler.
That is the workflow vocal chops actually take: isolate, chop, play, sequence. Doing it in a DAW means four tools and an export between each. Here it is one tab.
What to expect
How clean the acapella is
The separation is Spleeter, Deezer's open-source model (MIT licence), running in your browser. It is good for sampling, remixing and practice. It is not a studio multitrack: in dense, loud passages you can hear faint traces of the music in the vocal, and reverb tails tend to stay with the instrumental.
Sparse mixes separate better than loud, limited masters. For chops that is usually fine — a short slice of a vocal hides what a full verse reveals.
No upload
The song stays on your device
Upload-based extractors send your file to a server and make you wait in a queue. Here only the model is downloaded, once; the separation runs in the tab. That also means no account, no file-size metering, and no daily cap — TAIL is free.
It needs memory: about 125 MB per minute of audio while it works, which is why the limit is ten minutes per song. A recent laptop or desktop browser handles that comfortably.
More TAIL tools
Everything else that runs in the same tab
Sample chopper
Slice the acapella into chops on the waveform, then play them on 16 pads.
Stem separator
Four stems instead of two: vocals, drums, bass and everything else.
Key changer
Move the acapella into your beat's key without changing its tempo.
Common questions about extracting acapellas
Drop the song on this page and press Extract acapella. The vocal comes back as its own file, next to the instrumental. Download either, chop the acapella straight away, or load both onto TAIL's pads.
Yes. No signup, no watermark, no daily limit. The separation runs on your own device, so there is no server cost to pass on.
No. The separation model is downloaded to your browser once (78 MB) and cached; the song is processed locally and never leaves the tab.
About a minute for a four-minute song on a normal laptop, after the one-time model download. No GPU is needed.
Ten minutes per song, because the whole song is held in memory while it is separated. A longer file is offered a split into parts under ten minutes.
Extracting a vocal does not change who owns it. A sample of someone else's recording needs clearing before you release it; for practice, study and private remixes that is not a concern. Your own songs and properly licensed material are always fine.
Spleeter, the open-source separation model published by Deezer under the MIT licence, converted to run in the browser with onnxruntime-web.