Double Tracking Vocals: How to Get a Tight, Wide Stack
Recording · 4 min read
Sing the line again instead of duplicating the file. Keep the double 3–6 dB under the lead, pan a mirrored pair 60–100% left and right, and shelf them up 1–2 dB above 10 kHz. The small timing and pitch differences between two real takes are the entire effect.
Double level
-3 to -6 dB
Under the lead · panned 60–100% · +1–2 dB above 10 kHz
The 60-second answer
Record the line twice more. Pan take A 70% left and take B 70% right, both about 4 dB under the lead. On the doubles bus: high-pass 120–150 Hz, cut 2 dB at 250–350 Hz, shelf +1.5 dB from 10 kHz. The lead stays mono and dead centre.
Why a duplicated file is not a double
Copy the vocal, pan the copy, and you have made nothing. Two identical waveforms sum to one waveform 6 dB louder, still stuck to the centre. Add a fixed delay and a detuner and you get a comb filter that holds the same shape all song, which is why doubler plugins sound like plugins. A real second take drifts: each syllable lands 5–40 ms off the first, each note sits a few cents away. That moving difference is what the ear reads as width.
Why the double is quieter and brighter
Two takes at the same level read as two people, and the words get harder to follow because the consonants no longer land together. Drop the double 3–6 dB and the brain stops counting voices and starts hearing texture. Brightness solves it from the other end: the lead owns 200 Hz to 4 kHz, so anything you add there is competition, while a shelf above 10 kHz sits over the sibilance band and most of the vowel energy.
The stack, step by step
| # | Do this |
|---|---|
| 01 |
Track it, do not clone it Sing the full line twice more on the same mic, same distance, same session. Comp each pass before you stack. Lead + 2 doubles |
| 02 |
Match the notes, calm the delivery Same words and rhythm, flatter vibrato, simpler runs. Two moving pitches beat against each other. Doubles: less vibrato |
| 03 |
Pan as a mirrored pair A one way, B the same amount the other way. Under 50% the pair collapses inward and just blurs the lead. A: 70% L · B: 70% R |
| 04 |
Set level with the lead muted, then unmute If you can pick out a second singer, come down 2 dB. -3 to -6 dB |
| 05 |
Bus them and EQ once High-pass 120–150 Hz, bell -2 dB at 250–350 Hz, shelf +1–2 dB from 10 kHz. One set of moves. HPF 120–150 Hz · -2 dB @ 300 Hz |
| 06 |
Compress the bus, not the takes 2–3 dB with a 20 ms attack keeps the pair moving as one object. 2–3 dB GR on the bus |
| 07 |
Check mono The doubles should just get quieter. Thin or hollow means phase trouble: nudge one take 10–20 ms. Mono: no hollowness |
The trick: edit the ends of the words, not the beginnings
Everyone lines up the front of each word, because the transient is easy to see and easy to snap. But the ear catches a sloppy stack at the END of the word: the S that hangs on 60 ms too long, the -ing that fades late, the T that lands after the lead's T. Zoom in on the last consonant of every phrase in the doubles and slide it so it stops on the same sample as the lead. On a four-line chorus that is ten minutes of work, and it is the difference between a stack that sounds tracked in a bedroom and one that sounds bought. There is a bonus: once the word endings match, the doubles take a full extra dB before they read as separate people, so you get width without paying for it in clarity.
Stop guessing the chain.
The Vocal Chain Bible is 88 real vocal chains from Grammy-winning engineers — every plugin, every setting, in order, with free alternatives for each one. 5+ years, 100+ interviews, one PDF.
Get the Vocal Chain Bible — $48 →Mistakes that kill it
- Duplicating the file. Identical waveforms sum to 6 dB, not to width. A detuner on top adds a static comb filter.
- Doubling everything. If the verse is stacked four deep too, the chorus stops being an event. Leave the verse single.
- Panning the lead. The lead stays mono in the centre. Width comes from the layers around it.
- Copying the lead's automation. Doubles usually need to come DOWN where the lead gets loud, not up with it.
Two extra takes cost four minutes. Nothing you can buy replaces what they do to a chorus.
Keep going
FAQ
How many doubles does a vocal need?
Two is the standard stack, one panned each way. Four builds a thicker wall, but past that the extra takes mostly smear consonants.
Should the double match the lead exactly?
Same notes and rhythm, calmer delivery. Match the words and pull the vibrato back so the two pitch movements do not fight.
Can I use a doubler plugin instead?
For a demo, yes. In a mix it reads as an effect because the offset never changes. Use real takes on anything people sing along to.
Who's behind this
Ivan Meshcheriakov · IvanFromRnD
5+ years ago I moved to LA with zero connections and started building from nothing — 14-hour days, meeting people and interning at studios. It took me 5+ years to meet and interview 100+ Grammy-winning engineers — then I turned it all into the Vocal Chain Bible.
Read my full story →