How to Remove Vocals From a Song

The short version

Open the Music Editor, upload your song, and select Multi-Stem Split.

Set the separation type to Vocals + Instrumental (2 stems) and start the job. This is the two-stem mode, and it is the right one for vocal removal. A full split works too, but it does more work than you need and gives you four files to recombine.

When it finishes you get two players. One is the vocal on its own. The other is the same song with the vocal taken out, which is your instrumental. Download either or both.

You get the instrumental and the acapella, not one or the other

Vocal removal tools are usually described as if you pick an output. That is not how the separation works. The model splits the mix into a vocal side and a non-vocal side in a single pass, so both halves exist the moment the job finishes.

That is worth knowing because it changes what one job is worth. If you came for a karaoke backing track, you also walk away with the isolated vocal, which is the raw material for a remix, a mashup, or a sampled hook. One job, one charge, both files.

If the isolated vocal is what you actually came for, read how to get an acapella from a song, which covers what makes a clean one and why the vocal side is the harder half.

Why some songs strip cleanly and others leave a ghost

The instrumental is the forgiving half. Whatever the model gets slightly wrong tends to sit in the frequency range the vocal used to occupy, and the rest of the mix masks a lot of it. That is why vocal removal usually sounds better than vocal isolation on the same track.

Reverb is what betrays a removal. A model can identify the sung voice, but the reverb tail of that voice is spread across the whole mix and reads partly as room. On a dry, close-miked pop vocal you will hear almost nothing left behind. On a big ballad drenched in plate reverb you will hear a faint halo where the voice was.

Backing vocals and vocal-like instruments are the other tripwire. Stacked harmonies often leave together with the lead, which is usually what you want for karaoke and not what you want for a remix. A saxophone or a lead synth sitting in the vocal range can lose some of its body to the vocal side.

If a removal leaves too much behind, the fix is rarely a second attempt at the same file. Try a version of the song with less reverb if you have one, or accept the instrumental as a bed and layer over it.

What people actually use the instrumental for

Karaoke and live performance. The instrumental becomes the backing track to sing over. This is the most forgiving use, since a live vocal on top covers any residue.

A bed for content. Instrumentals sit under video, podcast segments, and social clips without a competing voice in the mix.

Practice. Removing the vocal gives a singer the arrangement to rehearse against. Removing everything but the vocal gives them a reference to learn the phrasing from, which is the same job run once.

Remix and rework. Instrumental plus isolated vocal is the standard starting kit. From there, the editor can extend a section, swap the instrumental, or add a new vocal line over what you kept.

Formats, limits, and cost

Upload MP3, WAV, M4A, FLAC, or OGG, up to 500 MB. A compressed source works fine; the output simply cannot recover detail the compression already discarded.

Separation runs on CPU and takes several minutes on a full-length track, with a 30 minute ceiling per job. Both output files are written as MP3. WAV versions are available on the Studio plan and above, transcoded from those MP3s rather than recovered from a lossless original.

A job costs 15 credits and is free on the Artist plan and above, the same as any other separation. Current rates are on the pricing page.

Frequently Asked Questions

How do I remove the vocals from a song?

Upload the song to the AetherWave Music Editor, choose Multi-Stem Split, and set the separation type to Vocals + Instrumental. The job returns the instrumental and the isolated vocal as two separate files.

Does removing vocals also give me the acapella?

Yes. The model splits the mix into a vocal side and a non-vocal side in one pass, so both files come out of the same job at no extra cost.

Why can I still hear the singer faintly?

That is almost always reverb. The model can separate the sung voice, but the reverb tail of that voice is spread across the mix and partly reads as room sound, so it stays behind. Dry, close-miked vocals strip much more cleanly than heavily reverbed ones.

Will backing vocals be removed too?

Usually yes. Stacked harmonies tend to leave with the lead vocal, which is what you want for a karaoke track and not what you want if you were trying to keep them.

Is removing vocals from a song legal?

Separating audio does not grant you any rights to the underlying recording. Use your own material, or material you have permission to use. Personal practice is a different situation from publishing or distributing a remix, and the recording owner still controls the latter.

What does vocal removal cost on AetherWave?

It costs 15 credits per job and is free on the Artist plan and above, the same rate as any other stem separation.