This is the third in a series on answering the question of how to isolate vocals from the rest of a track. The series started in Part 1 with some background and plugins, then explained what kind of tracks are more suitable than others in Part 2 and now delves into the specifics on tools and techniques.
It is pretty rare that you will use just one of the techniques we cover. Depending on the source material you start with, you can expect to combine several of these techniques to get the sound you are looking for. Actually, let me clarify: the “sound you are looking for” is probably the vocals a capella which is not going to happen so let’s revise that to the “the best sound you can get considering the singer has been dead for 55 years and the only recording of her voice is intertwined with a great but irrelevant pianist smashing chords directly in the singer’s prime range, right over her most emotional moments.”
A word of warning to the acoustically sensitive: what follows is not a very technical discussion and will probably hurt your ears just to read what kind of things are done to manipulate the sounds. For the rest of us, one thing to keep in mind is that you are not likely to get a very “good” sound in the end of this without taking acoustics into a account. That level of refinement is beyond the scope of this article. (Some might argue it is beyond the scope of my brain, but that’s another discussion.)
- EQ Filtering
I could show you a chart of where male and female frequency ranges fall but you’re going to roll your mouse off the table yanking around on those virtual knobs anyway so really, why bother with numbers?
Do the obvious: lop off the bass, lop off the highs. The vocals will be somewhere in the middle. Understand that you do not have to get completely rid of all instruments to be effective. A certain amount of lingering notes, especially padded, non-rhythmic background is workable. In my previous Howlin’ Wolf example there are screeching lead guitar riffs all around the vocals that I had no choice but to work into the mix. It worked for what it is which is all that matters.
EQ filtering can be a lot of mad science. There are lots of good resources for learning real world techniques in isolation. I learned a lot from the Q10 Setup Guide (pdf) and you don’t even need Waves Q10 to make great use of it. (Although if your EQ doesn’t have those options you might think about finding one that does.) There is also the venerable ProRec series on EQ. For a review of high-end products you can browse through a recent Electronic Musician comparitive review.
When I told a buddy I was writing this series he said: “You just use EQ. Right?”
Well, yea, there are only two things that happen in music: you are either creating sound or filtering it and since we are working with pre-recorded music this is all about filtering. Right?
Not always. When you clamp down the EQ on the track in an attempt to isolate the vocals you may be stuck with unseemly overlap between the vocals and instruments you don’t want. At that point you may end up “thinning” the voice of it’s bottom to get rid of the overlap. Sometimes the remaining vocals are usable as is. Sometimes it sounds like the singer phoned in the part. And I mean that literally, it sounds like they are singing through a rusty copper wire phone line.
In order to make up for that it helps to have a synth or guitar handy. Doubling the vocals in a subtle mellow tone will add body back into the performance. It needs to be mixed in with some care so as to feel ‘present’ but not audible. Click on Vincent on the right to hear a 30 second clip of what it sounds like when I doubled Gemma Hayes’ vocal line with a guitar.
You should be careful to blend the double as much as possible. It helps to have a soft attack and blend the release into the tail of the vocal. (ADSR envelope tutorial here.)
A twist is to make the double a harmony part a third, or full sixth below the melody line. If you don’t know how to do this have a piano playing friend come over and play the line for you. I’m actually serious, it’s well worth the experiment. As I write this, there is an Ozone MIDI Controller for $5.00 on eBay so you are running out of excuses.
Actually, gating is usually done based on volume, but here we want to dampen the volume of the parts between the singer’s words and syllables.
We also want to lessen the impact of snare hits directly in the clip. There are gate plug-ins and effects but I recommend using one of those envelope drawing tools they have in Sonar, Sound Forge or other wave editing tools. show_clip2("Elvis","left","Elvis
In FL Studio you can approximate the envelope by recording the slider movements and then hand edit the results in the event window.
As an example I picked a particularly hostile audio clip. Elvis Costello does not sing way out in front of his rock ensembles as in this clip where he uses guitar and piano all over his range. Plus there’s a snare slapping away right on top of the vocals. You can click on the two Vincents to hear what the original clip sounds like and then after I’ve EQ’d and applied a manual envelope. The ‘after’ may sound weird but you can’t simply gate the vocal words without some trailing off. If you strip out the music 100% between words all the better (!) but I picked this clip precisely because it was hard to work with and therefore the drop outs are never total.
The image on the left is an approximation of what the envelope ended up looking like for part of the clip. If you don’t go through this you will find your remix constantly fighting for supremacy with the original instruments. If you’ve done any reharmonization or you are in a different tempo then it’s impossible to work without this step.
- The Chopping Method
If you are using small samples or phrases from the original then the most control you can get is to chop up the vocals. I usually do this after all the steps I have outlined above. When chopping up you should keep the following rule in mind: Get the vocal phrase as close to the beginning of the wav segment as possible. You want to get is so that as soon as you hit the key, the vocal will sound. This applies to whatever method you use to trigger the sample. If you have chopped up drum samples then the method of actually tearing the vocal pieces apart should be very familiar. (If you don’t know what a zero crossing is, now would be a good time to learn.) Here are some of the methods I’ve used or heard of in the past. As usual, no one solution will work for all cases:
- I have spent a fair amount of time in my life splitting waves in ACID and dragging them into place over remix tracks. The advantage of this method is that you get a huge amount of control and the process has fantastic visual feedback. Newer tools like Tracktion extend the UI metaphor by allowing you to drag/drop effects onto individual clips. But I have abandoned this technique altogether because it is authoring and being a musician I like to perform my performance, not author it. The remaining methods are focused on ways to assign snippets of the vocals to the keyboard.
- Chop the sample into one-shots. Sound Forge’s “Extract Regions…” feature is very handy for this, as is ACID “Chopper” tool. Then assign each to a MIDI trigger in your host application. A good software sampler like Virtual Sampler or drum samplers like Native Instrument’s Battery were designed for just this kind of operation. It is also very easy in mixer hosts such as Ableton’s Live and FL Studio using layers although in both cases it gets tedious and overwhelming if you have chopped the vocals into many tiny snippets. If you choose this method, remember to route all channels to a single bus so that you can control all the vocals with one effects chain.
- Create a SoundFont using a tool like Vienna or MAZ’s Virtual Sampler. If your mixer can host VST/VSTi plug-ins then you can use the rgc free plugin for playback. The downside to this method is the amount of effort it takes to create a SoundFont is rarely worth it when you are going to use the vocal sample for just one remix and then toss the SF2 file. If you’re one of those guys or girls that can whip out an SF2 (and MAZ makes it about as easy as you can get) then this might work the best for you.
Use a ‘beat-slicer’ tool like ReCycle or ImageLine’s Beat Slicer that comes with FL Studio. This the easiest method I know of because it does the chopping and keyboard assigning in one step. They also have the advantage of not physically chopping up your wav but work with control (meta) data. These tools try to automatically calculate where the ‘beats’ are and for this application you need to completely ignore these ‘guesses’ and mark the snippets yourself which is usually no more than dragging some markers around and listening closely.
You can click on Vincent to the right to hear a quickie remix of the Elvis clip above as I “played” the samples in FL Studio.
There are many tools for chopping waves and I’ve just mentioned a few that have worked (with mixed results) for me. Many loop-heads that I know swear by Wave Surgeon for doing everything and more that I mention above. Unfortunately I have not tried it recently. If any readers have experience with it (good or bad) feel free to pipe up and let the rest of know about it.
That should be plenty for now but in the next edition of this I’ll be covering more even more techniques like non-chopping, the ever critical tempo-based-delay and several others. I know many of you have specific questions about specific applications and that’s fine, ask away. The worst thing that can happen is that I humiliate myself and it goes into the Internet Archive WayBack machine forever and ever. Oops! Already there…