Source separation
Source separation splits mixed audio into components — vocals from music, one speaker from another — using neural models trained on the task.
'Music/vocal split' became consumer-grade, and speech separation improves transcription of cross-talk. Separation quality degrades with similarity: two similar voices are far harder than voice-versus-music.
Related terms
Frequently asked
What is Source separation?
Source separation splits mixed audio into components — vocals from music, one speaker from another — using neural models trained on the task.
Why does Source separation matter?
'Music/vocal split' became consumer-grade, and speech separation improves transcription of cross-talk. Separation quality degrades with similarity: two similar voices are far harder than voice-versus-music.
What terms are related to Source separation?
Closely related concepts: Speech enhancement, Speaker diarization — each has its own entry in this glossary.