Vocal Remover
Split a song into its vocals and its music, or into vocals, drums, bass and the rest, with an AI model, without uploading it: karaoke, a cappella, practice
Drop a song here or click to browse
MP3, M4A, WAV, FLAC, OGG, or a video. Nothing is uploaded.
Song
Stems
Separate
An AI model (HT Demucs, by Meta) hears the voice, the drums and the bass apart, on your device. It is downloaded the first time (172 MB), then kept in your browser until you forget it.
It takes about as long as the song on a recent computer, and needs 2 GB of memory: a computer rather than a phone.