Remove Vocals From a Song, Free
Drop an MP3, WAV, or MP4 and get the instrumental and the vocal as two separate files — processed on your own machine, not on a server.
67 MB model downloads once, then stays cached.
Drop an audio or video file here, or click to browse
Audio or video · unlimited runs · no length or size limit
Desktop only · WebGPU required · Free and unlimited · Audio never leaves your device
How it works
- 1
Drop the file
MP3, WAV, M4A, and most video containers. Nothing is uploaded, so there is no size or length limit to hit.
- 2
Let the model run
A 67 MB vocal model loads once and runs on your GPU. A four-minute song usually takes one to two minutes.
- 3
Keep the instrumental
Download both stems as 192 kbps MP3. Use the instrumental for practice, the vocal for a remix, or both.
What a Vocal Remover Actually Does
A vocal remover estimates the singing in a mixed recording and subtracts it. The model works on a spectrogram: it finds the parts that look and sound like a lead vocal, then rebuilds the rest of the mix without them.
That means the instrumental you get is not a re-recording. It is your original file minus an estimate, so the drums, the bass, and the room are the ones you started with.
Vocals or Instrumental — Pick the Right Stem
Both files come out of the same pass, so you never run the song twice. The instrumental is the mix without the lead; the vocal stem is the estimate itself.
If you want an a cappella, take the vocal stem. If you want a backing track to sing over, take the instrumental. On dense mixes — distorted guitars, wall-of-sound production — expect some bleed in both directions.
Why It Runs Locally
Every free separation in this tool happens inside the browser tab. The file never travels to an upload endpoint, because the free tool does not have one.
The trade-off is speed: your GPU decides how long it takes. If a track is long or the machine is old, cloud separation runs the same model family on our hardware for a monthly fee.
Supported Files and Limits
Audio is decoded with the Web Audio API, so anything the browser can play can be separated. Video containers are decoded first and only their audio track is fed to the model.
- Audio: MP3, WAV, M4A, FLAC, OGG
- Video: MP4, MOV, MKV, AVI, WMV, FLV, TS
- Output: instrumental and vocals as 192 kbps MP3 at 44.1 kHz
- Runs: unlimited — no daily counter, no watermark
- Account: not required
Specs
| Input | MP3, WAV, M4A, FLAC, OGG, or a video file |
|---|---|
| Output | Two MP3s — instrumental + vocals |
| Runs on | Your device (WebGPU) |
| Upload | None on the free tool |
| Price | Free, unlimited |
Local or cloud
| Local (free) | Cloud | |
|---|---|---|
| Where it runs | Your browser (WebGPU) | Our servers |
| Track length | No limit (your machine decides) | 10 min (Starter) · 15 min (Creator) |
| Speed | Roughly 1× track length | Usually under a minute |
| Cost | Free, unlimited runs | From $2.99/month |
Frequently Asked Questions
Will the vocals be completely gone?
Usually not completely. The model leaves some backing vocals and reverb tails in the instrumental — that is how neural separation behaves, not a setting you can switch off.
Is there a file size or length limit?
Not on the free tool. Separation runs on your hardware, so the practical limit is your machine's memory and your patience.
Does it work on a phone?
The free tool needs desktop WebGPU. On mobile, use cloud separation, which runs on our servers instead.
Is my song uploaded?
No. The free tool has no upload endpoint; audio is decoded and processed in the tab. Cloud separation does upload, and deletes the input as soon as the job finishes.
Can I use the result commercially?
Free runs are for personal use. Cloud plans include commercial use of the stems you generate, provided you hold the rights to the original recording.