Have you ever wondered how to separate a vocal or the instrumental backing from a stereo mix? We find out what’s possible — and what isn’t.
The audio files on this page accompany the ‘Splendid Isolation’ article on vocal extraction/isolation, which appears in SOS June 2016. The first two are the original mixes, which I used in all the examples that follow:
My ‘sparse mix’ for Cristina Vane’s ‘So Easy’. The minimal arrangement makes it a better candidate for most vocal-extraction techniques.
My ‘full mix’ for Cristina Vane’s ‘So Easy’. The more complex mix makes isolation of the vocal from the instrumental backing more difficult in almost all cases.
These next two are based on the Mid-Sides approach:
With the sparse mix, the Mid-Sides approach does a reasonable job of removing the vocal.
As expected, though, the same approach is less successful when used on the full mix. Still, it might be acceptable enough for a bit of casual karaoke (if karaoke is ever acceptable!).
These next two are based on ‘Going Solo’ section of the article:
This is the full instrumental mix that I created specifically for this article. It should be identical to the CV So Easy full mix but without the vocal.
The results of using the phase cancellation approach with the ‘full mix’ versus the ‘full instrumental mix’ approach. As expected, the vocal can be isolated very cleanly.
This is also based upon the Going Solo section but uses the instrument track created via the mid-side process and tries to use that to ‘cancel’ the backing in the full mix and obviously fails badly J
As explained in the article, this example is included for completeness; as you can hear, the results are poor. A backing track was separated out using the M-S approach, and cancelled against the main mix in an attempt to isolate the vocal.
The next two examples result from processing using Melodyne Editor:
Used with the sparse mix, Melodyne allows a reasonable stab at the vocal isolation. It’s far from perfect but, if the isolated vocal was to be placed within a new musical context, it might well be good enough.
Again, as expected, attempting the same isolation from the full mix proves to be much more challenging.
These next on was created using Sony’s SpectraLayers:
Isolation of the vocal from the full mix using SpectraLayers. The results during the verse section are not too bad but for the busier chorus section, things proved more difficult.
These next four were created using Audionamix ADX Trax Pro. I did the two myself; the last two were done by the Audionamix team, to show you what difference a bit of experience using their software can make!
Separation of the lead vocal from the sparse mix using Trax Pro as described in the main text. Placed back into a busy re-mix context, much of this might be quite useable.
Separation of the lead vocal from the full mix using Trax Pro as described in the main text. The isolation is less complete and required more work to achieve. With more experience using Trax Pro, however, I’m sure the results could be improved further.
Separation of the lead vocal from the sparse mix using Trax Pro but performed by staff at Audionamix as a comparison with my isolation attempt. As might be expected, this is an improvement over my own effort.
Separation of the lead vocal from the full mix using Trax Pro but performed by staff at Audionamix as a comparison with my isolation attempt. While this is better than my own attempt, there are still some audio artefacts. This demonstrates that even using what is probably the best software tool available for this task, it can be extremely difficult to extract a high quality isolated vocal from a stereo mix.