How to use
- Choose the input files or text.
- Adjust the options and start processing.
- Verify the result and download it.
Split a song into vocals and accompaniment with an AI model and save each track.
Supported input is MP3, WAV, M4A, OGG. Speed and maximum size depend on device memory and the browser engine.
Primary flow checked in Chromium; other browsers and physical devices need separate verification.
Yes. Local tools never attach selected files to network requests.
Close other tabs, process smaller batches sequentially, and refresh the page to release memory.
Separated on your device with Deezer’s Spleeter (MIT) 2-stem model. About 80 MB downloads once; a 3–4 minute song takes tens of seconds on a GPU or a few minutes on CPU. Up to 15 minutes. Intended for personal practice; check the rights before redistributing.