She sang it six times. The first two were warm-ups — pitch cautious, phrasing polite, the kind of singing that happens when a voice is still negotiating with the microphone. Takes three and four were strong. Clean. The kind of performances that make an engineer nod and mark the session notes. Take five had a crack in the second verse, a moment where breath overtook intention, and the singer waved it off before the last chorus even landed. Take six was a victory lap — confident, locked in, almost too smooth.
They chose take five.
This is the counterintuitive work of comping vocals, where records are quietly won or lost. Comping — short for compositing — is the process of assembling a final vocal from multiple takes, sometimes syllable by syllable, sometimes in broad strokes. On paper, it sounds mechanical. In practice, it's one of the most emotionally demanding decisions in the entire recording process, because it forces you to answer a question with no clean answer: what does this song need from this voice, in this exact moment of this exact line?
The instinct is to reach for the take where everything went right. The pitch is centered, the timing tight, the consonants landing where they should. And sometimes that's the take. But more often than not, the performance that makes a listener lean in — the one that produces that physical thing in the chest — is the one with a seam showing. A breath that arrives a half-beat early because the singer was feeling the line before they were ready for it. A note that drifts flat at the end of a phrase because emotion was pulling harder than muscle memory. A moment where the voice does something the singer didn't plan, and the plan was worse.
The hardest part of comping isn't choosing between a good take and a bad one. It's choosing between a correct take and a true one.
The great producers know this. Rick Rubin has talked about choosing takes based on what he calls 'the feeling of truth.' It's not about perfection. It's about whether the hair on your arm moves. Jimmy Iovine once described sitting with Tom Petty for hours, listening to nearly identical takes of the same line, waiting for the one where Petty stopped performing and started meaning it. The difference was sometimes a single word. The rest of the take could be identical. But that one word, delivered with a fraction more weight or a shade more air, changed what the song was about.
This is why comping can't be fully automated. A pitch correction algorithm can tell you which take is most in tune. A timing tool can tell you which is most rhythmically precise. But neither can tell you which take has the thing — that unnameable quality where a human voice stops being a sound and starts being a confession. The tools are useful. They are not wise.
The correct take will survive scrutiny — headphones, studio monitors, the car. Nobody will hear it and think something is wrong. But the true take will do something the correct take can't. It will make the listener forget they're listening. It will make the song feel like it's happening for the first time, on the hundredth play.
The singer who waved off take five didn't hear what the room heard. She heard the crack, the breath that came too soon. What the room heard was a woman who, for twelve seconds in the second verse, forgot she was being recorded and simply meant every word. That kind of moment doesn't happen on command. You can't schedule it. You can only be rolling when it arrives, and then be brave enough not to erase it.
The best records are full of takes that almost got thrown away. The best engineers are the ones who fished them back out.