Can you remove vocals from a song for free?

Yes. ANYANO's vocal remover is free, needs no account and runs an open-source separation model inside your browser, so there is no upload, no queue and no watermark on the result.

Is the file uploaded anywhere?

No. The audio is decoded and separated in your browser tab. You can verify it in the network panel: processing a track sends no request carrying audio, and nothing is stored on our side.

How long does it take?

The first run also downloads the model, so it is slower. After that, expect roughly real time to a few minutes per track on a normal laptop. Keep the tab in the foreground while it runs.

Why can I still hear the singer faintly?

Because the model is reconstructing an instrumental that never existed as a separate recording. Reverb tails, doubled vocals, backing harmonies and vocal-range synths are the usual causes of residual bleed.

Can I release a track made from a removed vocal?

Separation changes the audio, not the ownership. Using someone else's recording still needs their permission, whatever tool produced the instrumental. This is general information, not legal advice.

Does it work on a phone?

Sometimes, on short tracks. Separation is memory-hungry, so a desktop or laptop is much more reliable, especially for full-length songs.

Guide

How to remove vocals from a song (free, in your browser)

A step-by-step walkthrough using ANYANO's free separation tools: no upload, no account, no watermark. Including the part most guides skip — what the result will and will not sound like.

Skip ahead to the tool

Separation runs entirely on your own machine using an open-source model. Your file is never uploaded, never stored and never used to train anything.

Open the AI Vocal Remover Or split all four stems

Walkthrough

Remove vocals in four steps

01
Open the Vocal Remover and pick your file
Go to the AI Vocal Remover and drop in an MP3, WAV, FLAC, M4A, AAC or OGG file up to 100 MB and eight minutes long. Nothing leaves your device — the file is decoded by your browser, not sent to a server.
Use the highest-quality copy you have. A 128 kbps MP3 ripped from a video will separate noticeably worse than the original WAV.
02
Let the model run locally
The first run downloads the separation model into your browser cache, then processing happens on your CPU. Expect roughly real time to a few minutes per track depending on your machine, and keep the tab open while it works.
A desktop or laptop is strongly preferred. Phones can run out of memory on long tracks.
03
Audition both results before you download
You get two results: the instrumental and the isolated vocal. Play them back to back and listen specifically to the loud chorus and to any quiet, reverb-heavy section — that is where bleed shows up first.
04
Download the instrumental you need
Export as WAV for editing or MP3 for sharing. If you plan to use the instrumental in a release, the recording still belongs to whoever owns it — separation changes the audio, not the rights.

The AI Vocal Remover before you choose a file: separation happens in the browser tab, and the limits are stated up front.

Quality

What the result actually sounds like

Separation quality depends far more on the mix you feed it than on which tool you use.

Usually clean

Modern pop, rap and dance mixes

A single dry lead vocal over a busy production separates well. The instrumental is normally usable straight away, with only faint artefacts on sibilance.

Mixed results

Heavy reverb, delays and doubled vocals

Long reverb tails belong acoustically to the vocal, so they either follow it out and leave the room sounding dry, or stay behind as ghostly wash.

Hardest cases

Live recordings, choirs, lo-fi and vocal-range synths

Bleed between microphones, stacked harmonies and synths that sit in the vocal range all confuse the model. Expect audible damage rather than a clean split.

Before you rely on it

Honest limits

Separation is an estimate produced by a model, not an undo button for a mix. These are the constraints worth knowing before you build anything on top of the result.

The original stems are gone. Anything recovered is reconstructed, so a perfect instrumental is not achievable from a stereo master.
Artefacts concentrate on cymbals, sibilance and reverb tails, and get louder the more you compress or limit the result afterwards.
Processing runs on your CPU, so long tracks take minutes and mobile devices may fail on anything over a few minutes.
Low-bitrate sources separate worse. Start from the best master you have access to.
Rights do not change. Removing a vocal from someone else's recording does not make the instrumental yours to release.

None of this is a reason to avoid separation — it is a reason to audition the result before committing to it.

Pick one

Which tool for which job

All of these run the same local model with different outputs. Choose by what you need back.

Tool	Use it when
AI Vocal Remover	You want the instrumental and the vocal as two files.
AI Stem Splitter	You need vocals, drums, bass and other as four separate tracks.
Instrumental Maker	You only want the backing track, ready to download.
Vocal Extractor	You only want the isolated vocal, for practice or a remix.
Voice Isolator	You are cleaning speech or a take, not a released song.
Background Music Remover	You want speech kept and the music behind it reduced.

The AI Stem Splitter, for when you need drums, bass and other as separate tracks rather than one instrumental.

FAQ

Frequently asked questions

Train your sound. Generate your song.

Separation is useful, but it is still someone's finished record. If you want material that is genuinely yours, train a private Sound on your own catalog and generate from it.

Keep reading

AI Vocal RemoverInstrumental and vocal, separated in your browser.AI Stem SplitterVocals, drums, bass and other as four tracks.BPM & Key FinderTempo, key and Camelot codes for your new stems.AI music copyrightWho owns what when AI touches a recording.