Reduces vocals that are mixed dead-center using phase cancellation (subtracting left from right). Honest heads-up: this reduces centered vocals on real stereo mixes — it isn't AI source separation, so it won't perfectly isolate every voice, and it does nothing on mono files.
The Vocal Remover creates a karaoke or instrumental version of a stereo song by cancelling the centre of the stereo image, where lead vocals are normally mixed.
Because the same cancellation also affects bass, kick and snare, this tool lets you restrict it to a frequency range (default 120 Hz–9 kHz) and choose a strength — so the low end and cymbals stay — which is why results are usually better than a plain “phase-invert and sum” trick.
A stereo mix is Left = Mid + Side and Right = Mid − Side. Anything panned dead-centre lives in the Mid signal. The tool band-pass filters Mid between your low and high limits (4th-order Butterworth), scales it by the strength and subtracts it from both channels, giving L′ = L − k·BP(M), R′ = R − k·BP(M). At full strength and full bandwidth that is the classic centre-cancel, leaving only Side information.
Honest limits: reverb and stereo-recorded vocals are not centred and partly remain; instruments that are also centred are reduced; and a mono file contains no side information, so nothing can be removed.
Make a backing track at home.
Play along to a track without the singer.
Create instrumental beds for videos (where you hold the rights).
Get a rough instrumental to work with.
| Method | Mid-channel cancellation, band-limited |
|---|---|
| Default band | 120 Hz–9 kHz, strength 100% |
| Requires | Stereo file |
No. It is a signal-processing technique that works on centre-panned sounds; AI stem separation can do more but needs server-side models.
That depends on how you use the result and who owns the recording; use it for practice or on audio you have rights to.
It works by cancelling audio that's identical in both channels (typically the lead vocal). Anything else mixed dead-center — bass, kick drum, some reverb — cancels too, and any vocal that isn't perfectly centered won't fully cancel. True, clean vocal removal needs an AI source-separation model, which isn't what this technique does.
If the source is mono, or the vocal was recorded/mixed off-center, there's nothing for phase cancellation to act on. It works best on well-produced commercial stereo tracks.
No — this is a classic, decades-old signal-processing trick (sometimes called karaoke-mode), not machine learning. It's fast and free but lower quality than a trained separation model.
This tool runs entirely in your browser using the Web Audio API. Nothing is uploaded, stored or shared — there is no server to send your file to, so it works offline once the page has loaded.
Be the first to share your experience with this tool — your feedback helps others.