Isolating Vocals from Background Music (Part 2)


This is the second part in a series on how to extract vocals from a track in order to us it in your remix. The first part covered general background and some plug-ins and methods that don’t really do the job. This entry is about what a Virtual DJ should consider when picking a track she plans to strip down to the vocals.

Picking Source Material

It is very likely the clip you’d like to use is not usable. The vocals might be buried too deep into the mix and unless you actually want that crunching guitar and slamming snare in your remix there is nothing you can do. There’s a lot to consider when analyzing a candidate clip. Sometimes you will try to manipulate a clip for a long time before you realize you just can’t make it work. It happens. A lot.

Here’s a list I try to keep in mind as I’m scouting around:

  • The performance should be very soulful and heartfelt. The more emo in the original vocal performance the better. That will help make up for the signal mutilation you are about to do. Go for honest, heartbreaking and gut-wrenching. I recently did a remix using indie singer-songwriter Shannon Campbell (original here). With source material like that it’s up to you to screw it up.
  • Go for loud singers. Here’s a remix I did using Howlin’ Wolf. He’s so loud he howls. Get it? Which makes him a great candidate for remixing.
  • Go for sparse instrumentation. As sparse as you can get it. Trying to extract the vocals from on top of a whole Phil Spector choir and orchestra is not trivial (if not impossible) so look for the “quiet parts” where it is just a small combo accompanying the vocals, like the intro, the bridge, the breakdowns, etc. Choppy, rhythmic accompaniment is usually preferred to legato draw out thick chords. But watch out: some of the toughest recordings to work with are singers accompanied by just a guitar! (The next point explains.)
  • Try to avoid passages with accompanying instruments right in the vocalist’s range. This is usually acoustic guitar, snare, synth lines, low shaker and some piano. My most successful extractions have been with small combos of bass, light piano and drummer playing with brushes (or no drummer at all!). I’m not sure why, but singer songwriter types seem to love strumming and picking away on these wide-body, big range acoustic guitars right smack dab in the middle of their own vocal’s frequency range. Pianists are only a little better about spreading their hands. The worst situation is when you get great extraction except for one pluck of the G string that can’t be suppressed.
  • Don’t worry about instruments in end ranges. The low range (like bass and thumping bass drums) and really high (like shakers and flutes) are the first and easiest to go. Not a problem.
  • Women are easier to work with. I mean vocally. I mean, as a rule. I mean always. I mean DJs, virtual or otherwise, like women’s vocals and not just for hormonal reasons. Female vocals tend to “sit on top of the remix” a lot better than male vocals. As a rule. I don’t know why and I don’t think I want to say any more because I will step on a land mine. Forget I said anything. And I don’t know what percentage of DJs are gay. So that proves my point. Or proves something. Probably about me. OK, I’m done here.
  • Live recordings, especially older ones (recorded before 1975) or bootlegs are usually a great place to start. Many of these shows are recorded with one mic and that one mic would usually favors the vocals a lot. I used a less well known live rendition of B. B. King’s “The Thrill Is Gone” in this remix (B. B. starts at 2:40). I’ve had the same luck with live Hendrix bootlegs and several others. In these cases the instruments just fall away when you apply the processing techniques I’ll outline later.

    It is also interesting to note that the live performances tug at the reader’s memory in a unique way: they recognize the song and even the singer, but not the particular performance and that can give your mix a particular blend of uniqueness and recognition.

  • Look for vocals where the singer enunciates the words. It is much easier to work with “I’m going to see you when you get a job” than “Imgonnaseeyawennyagettajob.” The bigger the breaks between words the better. The best example of this (and many other things) is Nina Simone. The worst is when the words actually run together “When nothing new” becomes one word “Whennnnothinnnnew” and singers love it when they get a chance to do that. They look hard for situations where they can get away with that. It adds, you know, ‘flavor.’ It also makes it hard to rely on that section as good remix material.
  • Depending on the type of mix you are doing, you can safely ignore tempo as a criteria. Even a four on the floor house remix can be pulled off when you up the BPM by 10. (The average listener will assume you’ve “sped up” the original even when all you’ve done is add bass drum on all four so you might as well pump it up if you need to.) Sony’s ACID remains the king of BPM stretchers so if you think you’ll be doing this or at least want the option then grab the cheapest ACID you can find (I’ve seen them for $10 at Costco in the States).

    If you’re going for more non-traditional sampling remixes rest assured there are software tools that you’ll want to use that stretch and chop the clip anyway.

  • A melody with less ‘passing tones’ will be easier to manipulate and match up to music in other keys and genres.

    A touch of theory: The melody of most popular forms of music tends to be a series of extremely simple ‘fence post’ notes. ‘Passing tones’ is the term musicologists use to define the frilly, filler notes singers and soloists use between the ‘real’ melody notes when they get expressive. Charlie Parker, Stevie Wonder and Justin Timberlake use lots of passing notes to very different effect. Passing notes are important to this discussion because they end up tieing the melody to a specific key.

    This can be a drag when you have extracted the vocals and you are trying to match it up to other music with a different chord structure. Other no single passing note is worth throwing out a whole idea. Again, the tools can save you and if everything else is working you might be surprised how flexible the ear will be to reharmonizations.

In summary, the most dangerous thing is to assume some clip you want to use won’t work. Chances are very good that, in fact, it won’t work but you should not assume that. You need to pretend as if it will work and at least try to use it. Like all things that are worth it, you waste a lot of time up front and waste slightly less time when you’re an expert. At least, that’s been my experience.

Next time (in Part 3 here) we talk about specific tools and techniques to isolating the vocals.

3 thoughts on “Isolating Vocals from Background Music (Part 2)

  1. Pingback: Brokenlooppers

Comments are closed.