What is Stem Separation?
TL;DR: Stem separation (source separation) is AI technology that isolates individual instruments and vocals from a mixed audio recording. Modern AI models can extract vocals, drums, bass, and other instruments from any song.
Deep learning models trained on thousands of songs learn the acoustic patterns of different instruments. Given a mixed song, the AI predicts what each instrument contributed, producing separate tracks that can be independently manipulated.