Remove vocals from any song with AI
The AI runs on your own device (WebGPU): your song is never uploaded, there is no queue, and the vocals and instrumental land as separate tracks you can edit.
How it works
Drop your song
Any MP3, WAV, OGG or M4A. The waveform appears in seconds. Nothing is uploaded.
Hit “Separate vocals”
The neural network runs on your graphics chip with live progress. A 4-minute song takes well under a minute on a typical machine.
Edit and export
Vocals and instrumental arrive as separate tracks: mute one for karaoke, keep the other for an a cappella, or remix both.
Why Ressoara
No upload, no queue
Cloud vocal removers put your song in a server queue. Here the AI runs locally: private by design, and never busy.
Studio-grade model
Powered by an open-source AI model (MIT), the same family behind the best-known desktop separation tools.
Real tracks, not just files
The stems land in a multitrack editor: cut, fade, add effects and export. No second tool needed.
Frequently asked questions
Really. The AI model is downloaded to your browser once (~64 MB) and inference runs on your device via WebGPU. Your audio never leaves your computer.
A browser with WebGPU: recent Chrome, Edge or Firefox, or Safari 26+. If your browser doesn't support it, the editor tells you honestly instead of failing.
Yes: the separation produces both stems: mute the instrumental track and export to get the isolated vocals.