Three problems get confused for one here, and they need different tools. Timing, where the layers do not start together and no alignment plugin fixes it because you are aligning the wrong part of the word. Pitch, where two sustained notes a few cents apart beat against each other and no tuner removes it. And spectrum, where a double carries the same low mids and the same top as the lead, so the ear hears a second singer rather than a thicker one. Work out which of the three you have before opening anything, because the fix for one makes the other two worse.
How do you tell which of the three you have?
Solo the lead with one double, and listen to a single word rather than the section.
If the problem is at the front of the word, it is timing. If it appears on a held note and sounds like a slow wobble, it is pitch. If it is there the whole time and the double just sounds like a separate voice at the same volume, it is spectrum.
Do this on the worst word in the chorus, not on a word that works. Twenty seconds of this saves an hour of moving faders.
Why does alignment fail on consonants?
Because you and the software are looking at different parts of the word.
The transient your eye finds in the waveform is usually the consonant, and consonants have soft, variable attacks. The letter s can ramp up over 60 to 80 milliseconds, and the singer will not produce that ramp identically twice. So aligning the visible attacks lines up the parts nobody locks onto.
The ear binds layers on the vowel onset. That is where pitch, loudness and vowel identity all appear at once, and it is the moment the brain uses to decide whether two sounds are one event.
The practical version: zoom in, find where the vowel actually starts rather than where the word starts, and align that. Tools like VocAlign do this well when the takes are close and struggle when they are not, which is why a take that was sung differently rather than late needs to be re-sung rather than processed.
One more timing detail that gets missed. Word endings matter more than people expect on stacks, because a consonant released at three slightly different moments reads as sloppiness even when every attack is perfect. Line up the release of the final consonant by hand on the last word of each phrase.
Why does tuning everything to perfect pitch still sound wrong?
Because two notes that are close but not identical produce beating), and no tuner is looking for it.
The arithmetic is quick. A cent) is a hundredth of a semitone, so 10 cents at 440 Hz is roughly 2.5 Hz of difference. Two sustained notes 10 cents apart therefore produce about two and a half amplitude pulses per second. That is slow enough to hear as a wobble and fast enough to sound like a fault rather than like vibrato.
Tuning both layers hard to the grid removes it, and takes the life with it. What usually works better is deciding which layer is the reference and correcting only the other one toward it, so the stack keeps one performance's worth of movement instead of averaging two into something rigid.
The exception is unison doubles on long held notes. Those want tighter correction than a harmony line does, because there is no interval to hide inside.
Why does a double sound like a second person rather than a thicker lead?
Because it was recorded like a lead and left full range.
A backing part that carries the same 100 to 300 Hz body and the same 8 to 12 kHz air as the lead is a complete voice in its own right, and the ear separates complete voices. Take either of those away and it stops being a separate object and starts being width.
What tends to work: high-pass the stack at 150 to 200 Hz, and pull 1 to 2 dB off the top with a wide shelf. Neither move is audible on the stack in solo, which is the point. If the stack sounds slightly dull on its own and correct in the mix, the setting is right.
Compress the stack bus, not each layer. Compressing layers individually pushes each one toward the front independently, and you end up with four foreground voices instead of one background texture. One compressor across the whole stack, doing 2 to 3 dB, glues them into a single object.
What about panning and reverb?
Panning is the cheapest fix on this list and the most often misused.
Hard-panning a pair of doubles left and right works when the two takes are genuinely different performances. It fails on copies of the same take with a delay, because the ear reads that as one source in a strange room rather than as two singers.
For reverb, send the stack and the lead to the same space but give the stack more of it. Different reverbs on lead and backing put them in two rooms, and two rooms is exactly what you are trying to avoid.
In what order should you do all this?
Timing, then spectrum, then pitch, then level.
Timing first because pitch correction on a misaligned layer bakes in the misalignment. Spectrum second because the moment you high-pass and shelf, the pitch problem often turns out to be smaller than it sounded. Pitch third, and only as much as needed. Level last, because until the first three are done you are moving faders to solve problems that faders cannot solve.
That order sounds pedantic until the first time you tune a stack for an hour and then discover a high-pass filter would have done it.
Frequently asked
How many layers should a stack have? Fewer than most sessions contain. Four well-sung doubles beat twelve that were recorded to compensate for the four. Each extra layer adds masking in the same region and buys progressively less thickness.
Should backing vocals be sung by the lead singer? Usually yes for unison doubles, because the timbre matches and the words land identically. Harmony parts can benefit from a different voice, since the interval already separates them.
Do stacks need de-essing separately? Yes, and normally more than the lead does, because sibilance from four layers adds while the vowels partly cancel. De-ess the stack bus rather than each layer.
If your stacks are fighting the lead
Most of the time the fix is one of the four moves above and takes fifteen minutes, not a new plugin. If you have a chorus where the stacks refuse to sit, send the session and say which word is worst, and I will tell you which of the three problems it is before you spend a day on the wrong one.
Production, vocal work, mixing and mastering happen at North Catalina Street in Burbank. Full credit list at /credits, background and contact at /about.