More audio tools: Audio Editor · Clean Voice · Text to Voice · All audio tools
What is this tool?
Split a song into vocals and instrumental, or into four stems (drums, bass, other and vocals), with Meta's HTDemucs model running in your browser. Use it to make a karaoke track, practise along with an instrumental, or isolate an acapella. Play the stems together, mute any of them, and download each one as WAV or MP3, or all of them as a ZIP.
How to Use
- Drop a song or a video with music.
- Choose Vocals + Instrumental or 4 stems.
- Press Separate. The model downloads once (about 110 MB including the AI runtime) and is kept for next time.
- Listen, mute stems to check them, then download the ones you need.
Frequently Asked Questions
How long does it take?
With a graphics chip (WebGPU) it is usually about as fast as the song plays. On a processor alone it can take several times the song length, and phones are slower still, so separate a shorter section there.
Can I use the stems in my own release?
Only if you own the song or have permission. Separated stems are still the original artist's recording.
Is the song uploaded?
No. The model runs on your device, so the song never leaves it.
Other tools
There are more calculators, converters and editors on Wivrix. If something is missing, tell us.