Splitting Songs into Stems
How StemFM separates a track into vocal, drum, bass, and melody stems — and what to expect.
3 min read. Updated June 25, 2026
Every session begins by splitting a song into stems. Understanding how it works helps you get the cleanest results.
The four stems
- Vocals — lead and backing voices.
- Drums — kick, snare, hats, and percussion.
- Bass — low-end and bass lines.
- Melody — everything else, including synths, guitars, and other instruments.
How separation works
StemFM uses AI source separation to reconstruct each layer of a recording. The model listens to the full mix and predicts what each individual instrument sounded like before it was combined.
After the split
Once separation finishes, each stem becomes its own interactive layer. From here you can isolate, mute, loop, and change speed without ever leaving the session.
