SWARA SOUND

AI Vocal Remover

Split any song into a clean instrumental (karaoke) and the vocals on their own (acapella). The AI runs inside your browser - unlimited, and your song never leaves your device.

Free & unlimited No sign-up Nothing uploaded
Drop a song here
or click to choose · MP3, WAV, M4A, OGG, FLAC
Choose audio file

🔒 Your song is separated inside your browser - it's never uploaded. The open-source MDX-Net model (from Ultimate Vocal Remover) downloads once and is then cached.

How it works

1

Open a song

MP3, WAV, M4A, OGG or FLAC. It opens in your browser - nothing is uploaded.

2

Separate

The AI model downloads once, then splits the song on your own device - about a minute with a graphics card.

3

Play & save

Listen to the instrumental and the vocals, then save either as MP3 or WAV.

How to remove vocals from a song

Make a karaoke version (instrumental) or pull out the acapella (vocals only) from any song. This free AI vocal remover and acapella extractor runs entirely in your browser - no upload, no queue, no limit.

  1. Open the song

    Drop an MP3, WAV, M4A, OGG or FLAC onto the box, or click Choose audio file. Nothing is uploaded.

  2. Choose a model

    Auto picks the best your device can handle. Best quality (64 MB) is quickest with a graphics card; Fast (28 MB) works well on any computer.

  3. Separate the vocals

    Press Separate vocals. The AI model downloads once and is cached, then the song is split on your own device - around a minute for a 3-minute song with a graphics card.

  4. Listen to both parts

    Play the Instrumental (karaoke - everything but the voice) and the Vocals (acapella - just the voice). Click a waveform to jump.

  5. Download

    Save either part as MP3 or WAV.

Frequently asked questions

How good is the separation?

It uses MDX-Net models from Ultimate Vocal Remover, among the best open-source vocal separators. On typical pop and rock songs the instrumental is clean enough for karaoke and covers. Very reverb-heavy vocals or voices hidden under loud guitars can leave faint traces.

Why does it take a minute (or several)?

Separating a song is heavy AI work, and it happens on your own device rather than a server farm. With a graphics card (Chrome or Edge on most laptops and desktops) a 3-minute song takes around a minute. On the processor alone it takes several minutes - the Fast model helps there.

Is my song uploaded?

No. The file is decoded and separated inside your browser tab. That's also why it's free and unlimited - there's no server doing the work.

Can I split drums, bass and other instruments too?

Not yet - this tool splits vocals from everything else. A four-stem model is about three times bigger, so we've kept this one light.

Can I use the instrumental on YouTube or in a release?

Separating audio doesn't change who owns it - the song's copyright still belongs to its owners. For music you can freely use, make an original track in the Swara Sound studio.

Sing over a track that's yours

Generate an original song in the Swara Sound studio - lyrics, style and vocals of your choice - and keep every file.

Open the Studio →

More free tools

See all tools →