Free AI stem splitter for vocals, drums and bass

What a stem splitter does

A stem is one part of a song: the vocals, the drums, the bass or the remaining instruments. Studios keep stems separate while mixing, but the finished song is one stereo file. A stem splitter works backwards: an AI model listens to the finished mix and rebuilds the four parts, so you can listen to, mute or download each one.

This stem splitter uses Meta’s open-source Demucs model and runs entirely on your device. Your song is not uploaded, and the four stems together sound like the original.

How to split a song into stems

1. Choose a song

   Drop an audio file or a video into the box. MP3, WAV, FLAC, M4A and OGG all work.

2. Separate

   Press the button and wait while the AI works through the song. The model downloads once on the first run.

3. Solo and mute

   Switch stems on and off and change their volume. All stems plays the full song, rebuilt from its parts.

4. Download the stems

   Each stem has its own download button and saves a WAV file you can open in any audio editor or DJ app.

What people do with stems

• Practice an instrument – mute the bass or the drums and play that part yourself over the rest of the band.
• Remix and DJ – load the drum and vocal stems into your DAW for edits, mashups and live blends.
• Transcribe and learn – a solo bass or vocal line is far easier to write down by ear.
• Karaoke – keep drums, bass and other instruments, and drop the vocals.

Quality is best with studio recordings. Instruments that share a range, like a synth pad and a choir, can blend between the vocals and other stems.