SOFTWARE EVOLUTION / ON-DEVICE AI
Sorinel Stems
One track.
Four ways to hear it.
Vocals, drums, bass and everything in between. Our in-house separation model brings out the layers of supported tracks, right on Apple silicon.
Hear the separation01 / LISTEN CLOSELY
The difference is in the layers.
Choose a demo and switch between the original, vocals, drums, bass and other. Hear what each part brings to the track.
PLAY
02 / INSIDE THE MODEL
Meet SorinelStems
A proprietary neural network family for music stem separation.
SDR — Source to Distortion
SDR
Separation Quality (dB)
Per-stem SDR from our internal family-balanced evaluation of SorinelStems V6. Results naturally vary with genre, mix complexity, and source audio quality.
100% on-device Apple Silicon ~5.5× realtime on MPS
1 Model
~5.5x
44.1kHz
4 Stems