This is a short segment, about 10 seconds for demo only. You can hear the background music and the vivid voice of the vocalist, all studio mixed, nothing has been changed (beyond just extracting this segment from a CD I own, so WAV file had full fidelity).

The scope of this demo is to show an example of Blind Source Separation

Stereo_12 : a short segment from an Original track

A propriety approach developed by the author (me) was used to remove the vocals, this is a pure prototype in the sense that nothing is tweaked yet, which makes the results even more smashing !

This is a cutting edge demo of machine learning & signal processing.

The original music segment is this one : https://soundcloud.com/user-637335781/stereo_12-a-short-segment-from-an-original-track

P.S. 
Needless to say that no trivial approach such as invert and merged channels or band pass filtering was used.

Stereo 12e - Algorithmic processing for removal of vocals

Please listen to the original song titled Hello by Adele (for example on you-tube or your favorite streaming service), between the time range of ~18 seconds thru ~58 seconds, then listen to this version. which is generated by a machine.

LinkOne1