Why are some vocals not fully removed?

Center cancellation only removes what is identical in both channels. Stereo reverb, delays and double-tracked harmonies are not perfectly centered, so traces remain; live recordings and old mono or fake-stereo releases are hard to process. Studio-produced pop songs work best. For stem-quality separation you need a dedicated AI model.

Why is the instrumental mono, with weaker bass?

L−R leaves only the side signal, which is the same in both output channels, so it sounds mono. Bass and kick usually sit in the center and get reduced along with the voice. Enable Keep the bass to blend the original low end below the crossover (120 Hz by default, adjustable from 60 to 250 Hz) back in.

How usable is the extracted vocal?

It is a vocal-dominant approximation from frequency-domain center extraction and often contains some centered instruments. It is fine for reference, transcribing lyrics or learning a melody, but not a clean studio acapella.

Are my files uploaded? Is it free?

No upload — all processing runs in your browser via WebAssembly and is discarded when you close the page. The tool is free with no sign-up and no limits.

🎤

🎤Vocal Remover Online - Remove Vocals from a Song & Make Karaoke Instrumentals

Need a backing track to practice, record a cover or sing karaoke? This tool uses classic center-channel cancellation. In most pop mixes the lead vocal is panned dead center, so it is identical in the left and right channels; subtracting one from the other (L−R) cancels it, while guitars, keys and backing parts panned to the sides survive. Because centered bass and kick are weakened too, a “Keep the bass” option mixes the original low end below 120 Hz (adjustable) back in so the instrumental does not sound thin. You can also extract an approximate vocal track using FFmpeg’s frequency-domain center extraction (dialoguenhance). Preview results right on the page, normalize loudness, and download MP3 or WAV. To be clear, this is not AI source separation: it only works on stereo mixes with centered vocals, mono songs, live recordings and vocals drenched in stereo reverb give weaker results, and the instrumental comes out mono.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Vocal Remover

1Upload one or more songs (MP3, WAV, FLAC, M4A… or a video)
2Choose instrumental and/or vocals; toggle Keep the bass and Normalize loudness as needed
3Click Process, preview the results on the page and download MP3 or WAV

FAQ

Why are some vocals not fully removed?: Center cancellation only removes what is identical in both channels. Stereo reverb, delays and double-tracked harmonies are not perfectly centered, so traces remain; live recordings and old mono or fake-stereo releases are hard to process. Studio-produced pop songs work best. For stem-quality separation you need a dedicated AI model.
Why is the instrumental mono, with weaker bass?: L−R leaves only the side signal, which is the same in both output channels, so it sounds mono. Bass and kick usually sit in the center and get reduced along with the voice. Enable Keep the bass to blend the original low end below the crossover (120 Hz by default, adjustable from 60 to 250 Hz) back in.
How usable is the extracted vocal?: It is a vocal-dominant approximation from frequency-domain center extraction and often contains some centered instruments. It is fine for reference, transcribing lyrics or learning a melody, but not a clean studio acapella.
Are my files uploaded? Is it free?: No upload — all processing runs in your browser via WebAssembly and is discarded when you close the page. The tool is free with no sign-up and no limits.