How to remove vocals from a song you already own
Searchers usually have one file and one job: the voice should go, the band should stay. This page is that job. Drop the file, leave vocals muted, and download the mix once the AI pass finishes.
It is not a centre-cancel trick. AudioDrop splits the song into vocals, drums, bass and other, then mutes only the vocal stem. Kick, bass and guitars stay in the stereo image instead of collapsing into a thin karaoke cassette.
You need a file you can open. Bought downloads, rips you made, voice memos and bounces from a DAW all work. Spotify, Apple Music and YouTube do not give this page the audio, so save a copy you have the right to use.
What you should still hear after the vocal is gone
A good removal keeps the record's arrangement. Here is what usually remains, and what goes.
| Part | After removal | Why |
|---|---|---|
| Lead vocal | Muted | Its own stem, fader down |
| Backing vocals | Mostly gone | They travel with the vocal stem |
| Drums and bass | Kept | Separate stems, left in the mix |
| Guitars, keys, synths | Kept | They sit in the other stem |
| Vocal reverb tails | Sometimes a trace | Long tails are the hard case |
Remove vocals on iPhone, Android, Mac or Windows
On a laptop, open this page in Chrome, Edge, Firefox or Safari and drop the file. A full song is fastest there, because the AI model uses several CPU cores.
On iPhone, use Safari, tap the drop area and pick the song from Files. On Android, use Chrome and pick it from Downloads. The same mute-vocals preset loads. Phones just take longer, and very long mixes can run out of memory.
The first run needs the separation model, about 166 MB. On a phone or a slow connection you confirm that download first. After it is saved, the next song starts immediately even if you closed the tab.
A cleaner take than the old karaoke button
Phase cancellation removes whatever is in the centre of the stereo image. On a pop mix that is the singer, and also the kick and the bass. The sides stay, so the track sounds hollow and the vocal's reverb is often left behind.
Splitting first, then muting, leaves you a fader. Pull the vocal up a hair if you want a guide while you learn the melody, or leave it fully down for a backing track. When you want the opposite, the acapella extractor solos the voice instead.
When the last of the singer will not leave
Stacked harmonies, shouted ad libs and a vocal printed with a long hall are the usual leftovers. Distorted or lo-fi vocals are harder than a dry modern lead. Start from WAV, FLAC or a 320 kbps MP3 if you have the choice, and do not download the one-second preview. That preview is for checking you opened the right song.
If you are about to sing it, try the karaoke maker next and drop the key with the key changer. If you want every stem as its own file, use the stem splitter and export them one by one.