AI Stem Splitter — Split a Song into Tracks Online
The stem splitter takes a finished song apart into four tracks — vocals, drums, bass and the remaining instruments — with the Demucs network. There are two ways to run it: in this browser, on your own graphics card and without an account, or in the studio on our GPU server, which takes tracks up to 15 minutes, offers a six-stem mode and, by our measurement, pulls the vocal and the backing track noticeably cleaner. Input: MP3, WAV, FLAC, OGG or M4A. Finished stems download one by one or as a ZIP, and “Open in mixer” places them on four faders of the audio mixer — for balance, EQ, reverb and a single-file mix export.
How to split a song into stems
Drop an MP3, WAV, FLAC, OGG or M4A up to 10 minutes long. The cost is shown next to your balance.
The Demucs network on our server separates the track into vocals, drums, bass and instruments — usually in under a minute.
Each stem has its own player on the page; download them one at a time in FLAC or all at once as a ZIP.
One button — and the four tracks sit on the audio mixer’s faders: volume, pan, EQ, mix export.
Break a finished song into tracks with AI — from upload to mixer in a minute
Studio stem separation
Split the track into separate stems — vocals, drums, bass and instrumental — on our server with the Demucs neural network. Download each stem separately or all at once.
Up to 15 minutes and 160 MB per file.
The split above runs in your browser and needs no account. The studio runs on our server, so that one does need one — the file you chose is kept in this browser for half an hour, so you will not have to pick it again after signing in. Other things you can do without an account: vocals and the instrumental — Vocal Remover the song without drums — Drum Remover
In the mixer every track gets its own volume, pan, equaliser and compressor — and all of it lands in the export.
