Comping is an editorial job, not a technical one

Given ten vocal takes, the instinct is to find the best version of each line and assemble them. This reliably produces something technically superior and emotionally flat.

Why it fails

A performance has an arc. Each take has its own internal logic — where the singer pushes, where they hold back, what they are building toward. Cutting between takes cuts between arcs, and the result is a vocal where every line is excellent and the whole thing means nothing.

Listeners do not hear individual lines. They hear intent moving through a song.

Find the take that means it, then fix the three things that are actually wrong with it.

A method that keeps the arc

Work in two passes and keep them separate, because they use different parts of your judgement and doing them at once produces the flat result.

First pass: listen to every take all the way through, once, without touching anything. No notes on individual lines. You are answering one question — which take is the performance. Usually two or three are candidates and the rest are not, and you will know within a few bars.

Second pass: take the winner and identify what is genuinely wrong with it. Not what is less good than another take. Wrong — a flat note, a fluffed word, a breath in a bad place, a line that drops out. Usually there are three to six of them, and they are the only edits you make.

Everything else stays. The slightly weaker line in verse two stays, because it is part of the shape that makes the chorus work.

What actually justifies a cut

• A pitch problem you can hear on a first listen, not one you found while scrubbing.
• A timing error that pulls against the track rather than leaning on it.
• A word that is unintelligible.
• A physical noise in the wrong place — a breath cutting off a phrase, a chair, a rustle.
• A line the performer themselves flagged.

Notice what is not on the list: a line that is merely better elsewhere. That is the criterion that produces the composite with no centre, and it is the one that feels most like doing the job properly.

Cutting so it does not show

Cut on consonants, not vowels. A plosive or a sibilant masks a transition; a sustained vowel exposes it, because the tone and the vibrato will not match across the join.

Cut before breaths, and take the breath from the same take as the line that follows it. A breath belongs to the phrase after it, not the one before, and mismatched breaths are one of the most common reasons a comp sounds assembled without anyone being able to say why.

Watch the room tone. If takes were recorded hours apart, or the singer moved, the background between phrases changes at every join. Short crossfades on the silences fix most of it.

The check

When the comp is done, listen to it once from the top without looking at the screen. If you can hear the edits, they are wrong. If you find yourself listening to the edits rather than the song, you made too many.

Then the harder check: play the winning take unedited straight afterwards. More often than is comfortable, the raw take is better — not more correct, better — and the right move is to go back and fix only two of the six things you fixed.

If the comp is more accurate than the take and less convincing, you optimised the wrong variable.