Isolation of different human voices from audio

Separate two voices from a speech signal

Several years ago it was hard to extract voice from music and almost impossible to separate two human voices from mixed stereo or mono tracks. With AI algorithms' aid there are many services allowing separation of voice/music or even different kinds of instruments. Is is possible now to separate two human voice from one track now?

Or, two make things easier, if one singing (score provided) and one talking, how should it be done, manually or automatically, without losing too many overtones and reducing quality?

  • 1
    $\begingroup$ Maybe helpful: vocalremover.org $\endgroup$
    – S.H.W
    Jan 1, 2023 at 16:39
  • $\begingroup$ We can only speculate about how some AI or ML or whatever algorithm does to identify different quasi-periodic tracks that are added together. But once a pitch of a voice is correctly tracked, the pitched part of the vocal utterance can be isolated with a correctly tuned comb filter. $\endgroup$ Feb 3, 2023 at 1:43

1 Answer 1


DeMIX Pro Version 4.0 from AudioSourceRE can separate the lead vocal from not only the backing instrumentation, it can separate the lead vocal from backing vocals and from harmony vocals! Highly recommended!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.