The short answer
Open the Acapella Extractor, choose the song and let the AI separate it. The page starts on “Vocals only”, so you hear the isolated voice; “Download vocals” saves it as a stereo WAV file at 44.1 kHz and 16-bit, the same length as the song. The vocal stem contains the lead voice together with harmonies and backing vocals.
Open the free toolHow to isolate vocals from a song
Choose the best source
A studio version as WAV, FLAC or high-bitrate MP3 gives the cleanest acapella. Live versions bring crowd noise and room echo with them.
Separate the song
Choose the file. The first run downloads the AI model once (171 MB), then the song is split into vocals, drums, bass and other instruments, in parts of about eight seconds.
Listen critically
“Vocals only” plays the isolated voice. Listen with headphones, especially in the gaps between lines: that is where traces of drums or guitar are easiest to hear.
Download the acapella
Press “Download vocals”. The WAV file starts at exactly the same moment as the song, so it stays in time if you put it back over the original beat.
What ends up in the vocal stem
The model separates voices from instruments, not one voice from another. The vocal stem therefore holds everything sung: the lead, doubled takes, harmonies and backing vocals. Effects on the voice travel with it. Reverb and delay are part of the acapella, which usually sounds natural and is what most remixers expect.
Some things can end up on the wrong side. A choir mixed like a soft pad may land partly in other instruments. An instrument that sits in the vocal range and sings a melody, such as a lead synth, violin or saxophone, can leave a soft trace in the vocal stem. Spoken parts, shouts and ad-libs count as voice and stay in.
How clean can it get? In our development test, we mixed a computer voice with a generated beat and separated the result: the vocal stem matched the voice with a correlation of 0.99 and the beat with 0.00. That shows the separation works as intended on a simple mix. A real song with reverb, choirs and dense guitars is harder, so treat it as a check of the engine, not a promise for every track.
Common acapella problems and how to fix them
Most leftover sounds are quiet and sit between phrases. A few minutes in any audio editor removes most of them:
| What you hear | Likely cause | What helps |
|---|---|---|
| Faint hi-hat or snare between lines | Drum transients close to the voice | Short fades or cuts in the silent gaps between phrases |
| A low hum or bass rumble | Bass or kick energy that leaked through | A high-pass filter around 80–100 Hz; most voices have little below that |
| A soft guitar or synth line | An instrument in the vocal range | Accept it under a busy remix, or lower it with gentle EQ where it sits |
| A washy tail after each phrase | Reverb that belongs to the voice | Keep it for a natural sound, or shorten it with a gate or expander |
| Thin, slightly watery sound | A low-bitrate or heavily compressed source | Start again from a better copy of the song if you can |
Can you remove only the lead vocal?
Some people search for a lead vocal remover because they want the backing vocals to stay, for example to sing the lead themselves over the original choir. The Vocal Remover cannot do that: lead and backing voices share one stem, so removing the vocals removes both.
What you can do is keep the vocals stem at a low level under the instrumental. Then the choir is still audible, but so is a quiet version of the lead. For many karaoke and practice uses that is a good compromise, and “Download this mix” saves exactly that balance.
Using an acapella in a remix or for practice
- Remix: put the acapella on its own track at the very start, find the song’s tempo and build your beat on that grid. Because the stem has the full length of the song, every phrase sits where it did in the original.
- New tempo or key: stretching or pitch-shifting a voice is possible in most DAWs, but large changes add artifacts. Small changes sound far more natural.
- Singing practice: solo the vocals to learn the phrasing and breathing, then switch on the bass quietly as a pitch reference and sing along.
- Lyrics: words that drown in the full mix are often easy to understand in the acapella.
- Cover check: record your own version over the instrumental and compare it with the isolated original line by line.
Privacy, and what you may publish
The song is never uploaded. Your browser decodes it and runs the model with ONNX Runtime Web, on the graphics chip when WebGPU is available and on the processor otherwise. Only the model and the engine are downloaded.
An isolated vocal is still the singer’s performance and the songwriter’s work. Studying it, practicing with it and experimenting at home is fine. Releasing a remix or bootleg with the acapella needs permission for the recording and the song; derivative works are among the owner’s exclusive rights listed by the U.S. Copyright Office. This is not legal advice.
Questions you might still have
Is the acapella extractor free?
Yes, with no account, no daily limit and no watermark. The site is funded by ads around the tools, never inside your files.
Are backing vocals included in the acapella?
Yes. The vocal stem holds the lead together with harmonies and backing vocals. Choirs that blend into the arrangement can partly land in other instruments.
Why can I hear drums in my acapella?
Loud drum hits close to the voice can leave short traces, mostly in the gaps between lines. Short fades or cuts in those gaps remove most of them.
Can I isolate vocals from a video?
Yes, from video files on your device. The tool reads the audio of MP4, MOV, WebM and MKV files and saves the vocals as WAV.
What format is the acapella?
A stereo WAV file at 44.1 kHz and 16-bit, the same length as the song, so it stays in time when you place it over the original beat.
Sources & useful details
- Meta Research: Demucs music source separation (GitHub)
- Rouard, Massa, Défossez: Hybrid Transformers for Music Source Separation (arXiv 2211.08553)
- ONNX Runtime: how to add machine learning to your web application
- MDN Web Docs: Web Audio API
- U.S. Copyright Office: What is copyright?
These guides are written from the current LoschaZ tools and the sources listed above. Examples are illustrations, not tests on every device or promises of an identical result.
