AI vs on-device separation
When to use each, and what quality to expect.
Runs on your device
The vocal remover also has an on-device option that works without any server. It uses phase cancellation: sounds panned to the centre (usually the lead vocal, bass and kick) are separated from sounds panned to the sides. It's instant and private, but rougher — centre-panned instruments go with the vocal.
AI separation is far cleaner because the model has learnt what each instrument sounds like. Expect a little “bleed” on dense mixes and on instruments that share a range (for example piano and guitar).
