Vocals apart, music apart — right in your browser
Track
Drop a song here
Neural network
This is not fast: a three-minute song takes several minutes of work, and the tab should stay open the whole time.
For personal practice. To publish a backing track of someone else's song, you need permission from the rights holder.
💡 How this works
This is not channel subtraction. The old “remove the centre” trick subtracts the right channel from the left: the voice does get quieter, but the bass and kick disappear with it, and on mono or modern mixes it does not work at all.
Here a neural network does the work. The MDX-Net model from the Ultimate Vocal Remover project looks at the spectrogram of the track and extracts the music without the voice. The vocals are the difference between the original and the instrumental.
It will not be perfect. On a dense mix, traces of the voice stay in the instrumental and hints of music stay in the vocals. That is how such models are, not a fault of this tool.
The result is a WAV. Compressing after separation would add its own artefacts on top of the model’s. If you need an MP3, there is an audio converter.
Is it free to use?
Yes, this tool is free and needs no sign-up. Your file is processed right in your browser and is not uploaded to a server. There are no watermarks.