Remove vocals
Drop a stereo track to cancel its center-panned vocal and get an approximate instrumental MP3. This uses channel cancellation, not AI, so it works best on stereo mixes and may also thin out centered bass and drums. Everything runs in your browser, so the audio never leaves your device.
How center-channel cancellation works
In a typical stereo mix the lead vocal is panned dead-center, which means it sits at almost the same level in the left and right channels. Subtracting one channel from the other cancels everything the two channels share, including that centered vocal, while the instruments that are spread across the stereo field survive. The exact operation is ffmpeg's well-known karaoke pan: pan=stereo|c0=c0-c1|c1=c1-c0. The new left is the old left minus the old right, and the new right is the old right minus the old left.
This is not AI stem separation
There is no model here, just channel math. That has real consequences. It only does anything on a genuine stereo source. A mono file has identical channels, so subtracting them leaves near-silence. It also removes other centered content. Bass, kick, and snare are usually panned to the center too, so the instrumental can sound thin or lose its low end. And vocals recorded with stereo width, doubling, or heavy reverb are only partly cancelled. Treat the result as a quick, approximate instrumental, not a clean studio stem.
Is it private?
Yes. The processing is a WebAssembly audio engine running on your own machine. Nothing is uploaded, stored, or seen by us. There is no server to send files to, and it still works offline once the page has loaded.
Frequently asked questions
- How does this remove vocals — is it AI?
- No, there is no AI or machine-learning model involved. It uses center-channel cancellation: in most stereo mixes the lead vocal is panned dead-center, so it is nearly identical in the left and right channels. Subtracting one channel from the other cancels whatever is common to both, which removes the centered vocal and leaves the stereo-spread instruments.
- Will it work on any song?
- Only on true stereo tracks where the vocal is mixed in the center. A mono file has nothing to subtract, so it will come out silent or nearly silent. Vocals with heavy stereo reverb, doubling, or wide panning are only partly removed.
- Why does the bass or drums sound thinner afterward?
- Because bass, kick, and snare are usually panned to the center too, the cancellation removes them along with the vocal. That is the trade-off of this technique. It cannot tell a centered vocal apart from a centered instrument. The result is an approximate instrumental, not a clean studio stem.
- What format do I get back?
- An MP3 encoded at 192 kbps, named with an "-instrumental" suffix so it is easy to tell apart from your source. The input can be MP3, WAV, M4A, AAC, FLAC, OGG, Opus, or AIFF.
- Are my files uploaded to a server?
- No. The processing runs on your device using a WebAssembly audio engine. Your audio is never sent anywhere, and the tool keeps working even if you go offline after the page loads.
- Is it free, and do I need an account?
- It is free with no watermarks, no daily caps, and no sign-up. Drop a stereo file and download the result.