Acapella Extractor: Get the Vocals From Any Song
An acapella extractor separates the vocals from a finished song and saves them as their own file. This free acapella extractor isolates the voice with Meta’s open-source HTDemucs AI model, right in your browser. Drop in an MP3 or a music video and download the acapella as a WAV.
How to get an acapella from a song
An acapella (spelled a cappella in the dictionary) is a vocal with no instruments behind it. You can get one from a song in four steps, and all you need is the song saved as a file on your device.
1. Add the file
MP3, WAV, FLAC and M4A work best, and MP4, MOV and WebM videos work too, up to 8 minutes and 150 MB.
2. Pick an engine
AI separation gives the cleanest vocal stem, and Quick finishes in seconds.
3. Check the result
Play the acapella next to the original.
4. Download the WAV
Save the acapella, and click “Also download the instrumental” if you want the music too.
AI acapella extractor or Quick: which gives the cleaner vocal?
AI separation gives the cleaner, fuller acapella on most tracks, and Quick gives you a fast lead vocal from the center of the mix.
If a Quick result isn’t clean enough, the tool offers to rerun it with AI separation.
| Feature | AI separation | Quick |
|---|---|---|
| How it works | Splits the track into vocals, drums, bass and other parts, then keeps the vocal stem | Keeps the center of the stereo mix, where most lead vocals sit |
| What ends up in the acapella | Usually all the singing, including harmonies and backing vocals | Mostly the lead vocal; wide backing vocals stay with the music |
| Acapella file | Stereo WAV | Mono WAV |
| Speed in our test (20-second clip) | About 3 minutes | About 5 seconds |
| Longest file | 6 minutes | 8 minutes |
| Clean-up controls | None | Clean-up (Gentle, Balanced, Tight) and Cut low rumble |
| Works best with | A browser that can use your graphics card | Any modern browser |
What do Clean-up and Cut low rumble do?
These controls appear only after a Quick run. Clean-up sets how hard the tool works to isolate the voice, with Balanced as the default. Tight removes more of the band but can thin out breaths and reverb, while Gentle keeps more of both. Cut low rumble, on by default, trims deep bass that leaks into the vocal.
You get the instrumental too
Every run also builds the instrumental, so one pass gives you both stems for a remix or a backing track. If the music is your main goal, start with the Instrumental Maker.
Your song stays on your device
Many online acapella extractors upload your track to a server. This one runs the AI model on your own computer, so unreleased demos and client files never leave your machine.
On your first AI run, your browser downloads the model once (about 180 MB) and caches it. AI separation runs best on your graphics card through WebGPU. Google’s web.dev team reports WebGPU support in Chrome and Edge 113 or later on Windows, macOS and ChromeOS, Firefox 141 or later on Windows, and Safari 26 (source).
How to get the cleanest acapella
The source file shapes extraction quality more than any setting.
- Start with a studio recording. Live versions blend crowd noise and room sound into the voice.
- Use your best copy. A WAV or FLAC gives a cleaner acapella than a low-quality MP3.
- Expect a drier voice. Reverb spread wide in the mix tends to stay with the music, so add your own reverb when you mix.
- Pick AI separation for mono files. Quick needs a stereo mix to find the center.
Dense choruses can leave faint traces of the band, which an EQ cut or noise gate can reduce.
Using your acapella in a remix
Any DAW opens the acapella file, which came out as 16-bit, 44.1 kHz audio in our AI test. The tool doesn’t detect tempo or key, so find those by ear or in your DAW. Then line up the vocal stem’s first downbeat with your beat.
Extracted acapellas also work for mashups, DJ edits, learning a vocal part or writing out lyrics.
Can you use an extracted acapella in a released remix?
You need permission first, because extracting a vocal doesn’t change who owns it. The U.S. Copyright Office explains that a song and its recording are separate works, and licenses generally must be obtained separately from the owners of each (Circular 73A). Clear the rights with the label or artist before your remix goes public.
How we built and tested this acapella extractor
The team behind VocalRemover builds and maintains this acapella extractor. It uses HTDemucs, an open-source model from Meta whose researchers report state-of-the-art results on the MUSDB music separation benchmark (paper).
We tested both engines in Chrome on a Windows laptop with a 20-second stereo test clip. With the model already cached, AI separation took about 3 minutes and saved a stereo acapella and instrumental. Quick took about 5 seconds and saved a mono acapella. Times vary with your device and file length.
More free audio tools
These VocalRemover tools also run in your browser:
- Vocal Remover splits a track into vocals and music.
- Karaoke Maker removes the lead singer and offers a guide vocal.
- Instrumental Maker keeps only the music.
- Background Music Remover cleans up speech in videos.
Acapella extractor FAQ
Is this acapella extractor free?
Yes. It’s free, with no signup and no watermark on your files. You need a modern browser and the song saved on your device. AI separation works best when your browser can use your graphics card, and Quick runs in any modern browser.
Is my song uploaded?
No. Your browser opens the file, runs the AI model and builds the acapella on your device, so the audio never reaches a server. The only download is the AI model, about 180 MB, which your browser fetches once and then caches.
Will the acapella include backing vocals?
Usually, yes, if you use AI separation, which keeps all the singing it finds, including harmonies and backing vocals. Quick mostly gives you the lead vocal from the center of the mix, while backing vocals panned wide stay with the music.
Is the acapella stereo or mono?
It depends on the engine. AI separation saves a stereo file, so stereo effects on the voice stay in place. Quick saves a mono file because it pulls the vocal from the center. You can pan or widen a mono vocal in your DAW.
Can I extract an acapella from a YouTube video?
You can’t paste a link, but you can use a saved video file. The tool doesn’t fetch videos from YouTube or other sites. It reads MP4, MOV and WebM files on your device and pulls out the audio before it separates the vocal.
Why does my acapella sound dry or thin?
Reverb spread wide in the mix tends to stay with the music, so the vocal sounds drier. In Quick mode, the Tight clean-up setting can also thin out breaths. Try Balanced or Gentle, or switch to AI separation, then add your own reverb.
Why is AI separation slow on my computer?
AI separation runs a large model on your graphics card, so it takes minutes rather than seconds. If your browser can’t use the graphics card, it gets very slow, and the tool recommends Quick instead. A browser with WebGPU support helps.
What's the difference between this and the Vocal Remover?
Both tools use the same engine but lead with different results. This page works as an acapella maker, so the acapella comes first and the instrumental is an optional extra. The Vocal Remover gives you both tracks side by side from the start.
Ready for your next remix?
Choose a song and get a clean acapella, right in your browser.
Last updated: November 2026