Remove vocals from any song. Free, in your browser.
Drop a song in, get the vocals and the instrumental back. Your song never leaves your device — the split happens on your computer, not on a server. No account, no limits, downloads included.
Got several songs? Batch vocal remover — drop in a whole folder and split them one by one.
Split a song
Idle — waiting for a songDrop your song here
or pick a file from your device
MP3 · WAV · FLAC · M4A · OGG · MP4 video · up to 10 minutes
A direct link to an MP3, WAV or FLAC file — for example a public Google Drive or Dropbox link. YouTube, Spotify and SoundCloud links don't work: those sites don't let a browser fetch the audio, so download the file first and drop it in.
Model BS PolarFormer · 103 MB · downloaded once, then cached in your browser
Click the waveform to jump to that part of the song.
Vocals
The isolated voice track.
Instrumental
The backing track, vocals removed.
Download your stems
Free, unlimited, no watermark. Stems are still the original artist's work — don't sell or re-upload them as your own.
Your song never leaves your browser. The file is decoded and split on your own device. We have no server that receives it, no account that stores it, and no way to hear it.
Why this one is actually free
Most vocal removers rent you time on their servers. This one runs on your computer instead, so there is nothing to meter — and nothing to charge for.
Free, unlimited
No minutes, no songs-per-month, no credits. Split as many songs as you like.
Nothing is uploaded
The song is decoded and separated in your browser with ONNX Runtime Web. No server ever sees it.
Downloads included
Take both stems home as WAV or MP3, or grab them together as a zip. Free tier, full files.
No sign-up, no watermark
No account, no email, no ads in the way. The stems are clean.
How it compares
Same job, different deal. Free-tier details as published by each service, checked November 2026.
| Free tier | Vocal Remover | LALAL.AI | Moises | vocalremover.org |
|---|---|---|---|---|
| Free allowance | Unlimited | 10 minutes | 5 songs a month | Free tier |
| Download your stems | Yes — WAV or MP3 | No — preview only | — | — |
| File size limit | No limit, up to 10-minute songs | — | — | About 10 MB |
| Where your song goes | Stays in your browser | Uploaded to their servers | Uploaded to their servers | Uploaded to their servers |
| Sign-up needed | No | — | — | — |
| Ads | None | — | — | Yes |
“—” means we don’t make a claim about that cell. Limits change over time — check each service’s own site before you rely on it.
How it works
Three steps, one page. Nothing to install.
Drop your song in
MP3, WAV, FLAC, M4A, OGG — or an MP4 video. Your file is decoded right in the browser.
Your device splits it
A 103 MB open model (BS PolarFormer) downloads once, then runs on your GPU or CPU. No upload, no queue.
Listen and download
Compare the original, vocals, and instrumental, tweak the mix, and download the stems as WAV or MP3.
Questions
Is it really free?
Yes. There is no server doing the work, so there is nothing to pay for. The model runs on your own device, and the site charges nothing for it. No account, no credits, no per-song fee.
Does my song get uploaded anywhere?
No. The file is decoded and separated inside your browser. It never leaves your device, and nothing is stored on a server. Close the tab and it is gone.
What formats can I upload and download?
Upload MP3, WAV, FLAC, M4A, OGG, or an MP4 video (the audio track is used). Download each stem as WAV or MP3, or both together as a zip. Songs up to 10 minutes long are supported.
Can I paste a link instead of uploading?
Yes, if it's a direct link to an audio file — an MP3, WAV or FLAC on a public server, or a public Google Drive / Dropbox link. Paste it in the link box above and it loads straight into the splitter. YouTube, Spotify and SoundCloud links don't work: those sites don't let a browser fetch the audio, so download the song as a file first and drop it in.
What model do you use?
BS PolarFormer — a 103 MB open model (FP16) built on a newer RoFormer-style design. It estimates the vocals directly, so the vocal stem comes straight from the model and the instrumental is what is left of the mix. That is why it sounds cleaner than the older karaoke models that worked the other way round. It still struggles on songs with heavy reverb, dense backing vocals, or mono recordings; on those, some vocal ghosting in the instrumental is normal, not a bug.
How good is the separation?
Good on most pop, rock, and electronic songs — BS PolarFormer is a well-tested open model. It struggles on songs with heavy reverb, dense backing vocals, or mono recordings; on those, some vocal ghosting in the instrumental is normal, not a bug. If a song is hard to split, that is the honest limit of the model, and it will sound that way on other tools too.
Can I use the stems commercially?
No. Splitting a song does not change who owns it. The vocals and instrumental are still the original artist’s work — use them to practice, sing along, or remix for yourself, but don’t sell them or upload them as your own.
What if my browser has no WebGPU?
The tool falls back to your CPU (WASM). It still works, but it is slower — a 3-minute song can take several minutes instead of under a minute. The progress bar shows a real time estimate while it runs, and you can cancel at any time.