Will the acapella include the backing vocals?

Yes. The model separates every voice it finds into the vocal stem, lead and harmonies together. If you want only the lead, you will need to work on the stem in your DAW; the separator cannot tell singers apart.

Why does the vocal sound a bit wet or echoey?

Because the reverb and delay were printed on the vocal in the original mix. The separator gives you the voice as it appears on the record, effects included. That is usually what you want for a mashup; a completely dry vocal only exists in the original session.

What sample rate and format should I keep the stem in?

WAV, at the sample rate of the source. Keep it lossless until the final bounce; every MP3 encode in between adds artifacts that stack up audibly on an exposed vocal.

Can I extract an acapella on my phone?

The browser separator runs on phones too, more slowly and with the file-size limits of a phone browser. For a full-length song a laptop with a graphics card is the practical choice.

Guides

How to Extract an Acapella from Any Song (for DJs, Remixes and Mashups)

September 10, 2026 · 4 min read

For most of DJ history an acapella was something you found: an official release, a DJ-pool download, a lucky rip. Now it is something you make. Source separation pulls the vocal out of a finished stereo mix well enough to build a mashup on, and the interesting work moves from finding the file to fitting it into your track.

Acapella, vocal stem, isolated vocal: the same thing?

Close enough. An acapella in the strict sense is the vocal as the label released it, mixed and processed exactly as on the record. A vocal stem is what a separator produces: an estimate of that vocal, lifted out of the mix, with the reverb and delay that were printed on it. For a mashup the difference is small; for a commercial remix, a label-supplied acapella is still cleaner, because nothing was estimated. The acapella extractor gives you the second kind in a few minutes, without a server.

Step 1: Extract the vocal

Open the acapella extractor and drop in the track. Use the studio version and the best file you have; a WAV or FLAC separates noticeably cleaner than a low-bitrate MP3.
Let the model run. It downloads once (about 170 MB) and then runs on your device through WebGPU or WebAssembly, so the file is never uploaded.
Download the vocal stem as WAV. Keep the instrumental from the same pass too; you may want to borrow a bar of it later, and it lines up with the vocal exactly.

Step 2: Know the key and the BPM before you touch anything

A mashup lives or dies on two numbers. Get them from the original track, not from the stem, because analysis tools read a full mix more reliably than a bare voice. Rekordbox, Serato and Traktor all show key and BPM after analysis; Mixed In Key is the usual choice if you want a second opinion on the key. Write both down for the acapella and for the instrumental you are pairing it with. Compatible keys (the same key, its relative major or minor, or a fifth away) will sit together; anything else needs pitch-shifting, and vocals tolerate about three semitones before they start to sound processed.

Step 3: Line it up

Drop the vocal stem into your DAW or DJ software on its own track. Trim the silence at the start so the first word sits on a bar line, then set the clip's tempo to the original song's BPM so warping works from the right number. In Ableton that is a warp marker on the first downbeat; in Serato and rekordbox it is a beatgrid. Time-stretching a vocal by up to about eight percent is transparent; beyond that, the consonants smear and the vibrato slows audibly, so pick an instrumental within that range or accept a few artifacts.

Step 4: Clean up what the model leaves behind

Even a good separation leaves traces of the mix in the vocal stem: a hint of hi-hat in the sibilance, a wash of reverb, a little bass under a low male voice. Three moves handle most of it.

A high-pass filter around 80 to 120 Hz removes bass bleed without touching the voice.
A gentle gate, or manual fades between phrases, cleans the gaps where residual instruments are most audible because nothing is masking them.
For a song that comes back phasey or hollow, the HD model on the HD Pack or Pro plan runs a larger vocal-specific network on a server GPU and usually does better on dense mixes.

The old ways of getting acapellas, and why they were a lottery

Official acapellas exist for a fraction of releases and usually for the singles a label wanted remixed. DJ pools carry more, at a subscription. The classic do-it-yourself method was phase inversion against the instrumental version: if you had the instrumental and the full mix from the same master, inverting one and summing the two left only the vocal. It worked when the masters were identical to the sample and failed, noisily, whenever they were not. The "acapella" uploads on video sites were often exactly that trick, with the failures included. Separation replaced all of this because it needs only the file everybody already has.

Using acapellas legally

A bootleg mashup for your own sets is a long-standing grey area that most rights holders tolerate. Releasing a remix, even for free, needs clearance from the owners of both the recording and the composition, and separation does not change who owns what. Some artists explicitly invite remixes and publish stems for it; when they do, use those, because nothing beats a stem that was never estimated.

Frequently asked questions

Will the acapella include the backing vocals?: Yes. The model separates every voice it finds into the vocal stem, lead and harmonies together. If you want only the lead, you will need to work on the stem in your DAW; the separator cannot tell singers apart.
Why does the vocal sound a bit wet or echoey?: Because the reverb and delay were printed on the vocal in the original mix. The separator gives you the voice as it appears on the record, effects included. That is usually what you want for a mashup; a completely dry vocal only exists in the original session.
What sample rate and format should I keep the stem in?: WAV, at the sample rate of the source. Keep it lossless until the final bounce; every MP3 encode in between adds artifacts that stack up audibly on an exposed vocal.
Can I extract an acapella on my phone?: The browser separator runs on phones too, more slowly and with the file-size limits of a phone browser. For a full-length song a laptop with a graphics card is the practical choice.