How to Comp Vocal Takes Like an Engineer
Recording · 5 min read
Record 4–6 full takes, then listen once through and rate every phrase 1 to 3 without editing anything. Assemble the winners, and make every cut on a consonant with a 10–20 ms crossfade. Never splice in the middle of a vowel — a vowel is a continuous waveform and the join will click or wobble however short the fade is.
Where every edit goes
ON THE consonant
10–20 ms crossfade · never inside a vowel
The 60-second answer
4–6 complete takes, same distance every time. One listening pass, rating each phrase 1 to 3 on a lyric sheet. Build the comp from the 3s. Cut at the start of a consonant, crossfade 10–20 ms, match clip gain across the joins to within 1 dB.
Why a vowel splice always shows
A sung vowel is a periodic waveform — a repeating cycle at the pitch of the note plus its harmonics. Cut it and join it to another take and you are joining two waveforms at different points in their cycle, at slightly different pitches. Even with a crossfade you get a momentary cancellation and a tiny pitch step in the middle of a sound the ear is actively tracking. A consonant is the opposite: an S, F, T or K is noise with no steady pitch to disagree about, and it masks the join completely.
Rate first, edit second
Comping while you listen means every decision is made against the take you just heard rather than against all of them. You end up with the last decent version of each line instead of the best one, and it takes three times as long because you switch between listening and mouse work every four seconds. A rating pass costs one playthrough, and the edit pass afterwards is mechanical.
The comping workflow, step by step
| # | Do this |
|---|---|
| 01 |
Record complete takes, not punches 4–6 passes top to bottom. Whole takes keep the energy consistent and give you real alternatives for every line. 4–6 full takes |
| 02 |
Keep the distance identical every take Tape on the floor. Two takes recorded 8 cm apart have different low end, and no crossfade hides that. Same mark, every take |
| 03 |
Rating pass: listen once, edit nothing Lyric sheet, one column per take, rate every phrase 1 to 3. Circle any single word that was exceptional. Rate 1–3 per phrase |
| 04 |
Assemble the 3s in order Build the whole comp before refining anything. A rough full comp tells you more than one perfect line. Rough comp first |
| 05 |
Cut at the front of a consonant Find the silence or noise burst just before the word and put the edit there. Never the middle of a held vowel. Crossfade 10–20 ms, equal power if offered. Consonant onset · 10–20 ms fade |
| 06 |
Match levels and breaths across the joins Clip gain the pieces to within about 1 dB, and keep the breath from the take you are cutting into, not the one you are leaving. Clip gain within 1 dB |
The trick: always cut on a consonant, never in a vowel
This one rule removes most of the audible edits in a comp, and once you know it you place every crossfade somewhere different. Zoom in on a phrase and look at the waveform: the vowels are the dense, regular, repeating sections and the consonants are the spiky or fuzzy bits between them. Every edit belongs in a fuzzy bit. The front of any word starting with S, F, T, K, P or CH is a free edit point, and so is the tiny gap before a hard consonant inside a word. If the two lines have no consonant near the transition, do not force it — go back to the previous consonant, even if that means taking an extra word from the other take. Always better to move the cut than to shorten the fade. Two extras: check every edit in mono, where phase problems from a vowel splice are much louder, and comp a doubled stack so every layer changes take at the same word.
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
- Comping while you listen. You end up with the most recent acceptable line, not the best one, and it takes three times longer.
- Crossfades of 1–2 ms. Too short to remove the discontinuity. You hear a tick on the join and blame the plugin chain.
- Forgetting the breaths. A comp with a breath at every edit and none anywhere else sounds like a machine assembled it.
- Comping each stack layer independently. The lead changes take at one word and the double at another, and the stack loses its lock.
Rate once, assemble fast, cut on consonants. A good comp is invisible, and every mixing decision after it gets easier because level and tone stop moving.
Keep going
FAQ
How many vocal takes should I record?
4–6 complete passes for a lead. Beyond that the performance usually gets worse, not better, and comping gets slower.
How long should a vocal crossfade be?
10–20 ms at a consonant. Shorter clicks, longer starts blurring the attack of the word.
Should I comp before or after tuning?
Comp first, always. Tuning a take you will replace is wasted work, and tuned pieces from different takes rarely join cleanly.
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 →