Unmix
Pull the singer out of a song, right in your browser. Drop in a track, pick up to 30 seconds, and get the voice and the backing as two separate files. No training data, no upload.
1Choose a song
Drop a song here or
MP3, WAV, M4A, OGG or FLAC: anything your browser can play2Pick an excerpt and separate
This file is mono. One of the three cues, “the singer is in the centre of the stereo image”, can’t help, so expect more of the backing in the voice.
Works best on a song with one lead singer mixed in the centre. A 20-second excerpt takes roughly 15–40 seconds on a laptop at the default setting.
3The parts
Voice
Download WAVAccompaniment
Download WAVOriginal excerpt
The 16 kHz excerpt the separation worked on.
Everything runs in your browser, in a background worker: the audio never leaves your computer. It runs the same solver as the Python version, with fewer iterations by default so it finishes faster; details.
How it’s built, results and limits: How it works →