Why “good take” is not a judgement
Ask an inexperienced producer about a vocal and you get an adjective: it is good, it is not quite there, it needs something. Ask someone who does this for a living and you get a location: the second verse drifts flat, the consonants disappear in the chorus, the last line of every phrase runs out of air.
The difference is not sensitivity, it is decomposition. Professionals do not evaluate a take as one thing, because “not quite there” gives you nothing to act on. They evaluate five, and each one implies a different remedy: a punch-in, a nudge on the grid, an EQ move, another take, or a different key.
What follows is that framework. Its value is not that it sounds rigorous, it is that at the end you know exactly what to do next.
1. Pitch accuracy
The first dimension, and the one everyone reaches for. Intonation is measurable in cents, where a cent is one hundredth of a semitone: a professional take typically averages 5 to 15 cents of deviation, beyond 25 the errors become audible, and 50 cents is a quarter tone.
Two figures matter. The average describes the take as a whole; the off-pitch moments describe the exceptions, and those are what listeners actually notice. A take averaging a respectable 12 cents can still contain four sustained notes sitting badly flat, because averages hide outliers.
Only sustained, steady notes should count. Scoops, bends and vibrato are phrasing, and treating them as errors would condemn most soul, blues and contemporary pop singing.
The fix: isolated notes are a punch-in or light correction. A take that drifts as a whole is a performance or a key problem. There is more detail in how to tell if your vocals are in tune.
2. Timing and groove
The common misunderstanding is that good timing means landing exactly on the beat. It does not. Sitting fractionally behind the beat gives a relaxed, weighty feel and is characteristic of soul and much of hip-hop. Sitting fractionally ahead gives urgency and drive, and is common in punk and uptempo pop. Both are correct, both are choices.
What is not a choice is inconsistency. A vocal that is behind in the verse, ahead in the pre-chorus and exactly on in the chorus does not read as expressive, it reads as unsteady, and it makes the whole arrangement feel like it is slipping.
There is an opposite failure too. A take that lands exactly on the grid for most of its duration stops sounding human. Above roughly 70 percent of the time perfectly on the grid is already unusual for a sung performance, and above 85 percent is very hard to achieve without editing. Perfect timing reads as programmed, not as skilled.
The fix: individual late or early entries can be nudged. A consistently drifting feel is a re-record, ideally against a click the singer can actually hear over the headphone mix.
3. Diction and intelligibility
Diction is the dimension listeners notice fastest, because the moment a word is unintelligible their attention leaves the song to work out what was said. It never returns to exactly where it was.
It has two halves that get confused. Performance diction is how clearly the singer articulates: consonants that get swallowed, word endings that vanish, vowels that shift under pressure. Mix diction is whether those consonants survive the arrangement, which is a masking problem in the 2 to 5 kHz range rather than a performance one.
Sibilance sits at the boundary. Excessive energy in the sibilant band is a recording and mixing issue rather than a technique issue, and it is worth treating: past a certain point the esses become aggressive enough that de-essing stops being optional.
The fix: if the words are unclear in solo, it is the take. If they are clear in solo and unclear in context, it is the mix, which is covered in why your vocals get lost in the mix.
4. Delivery and emotion
This is the dimension nobody can measure and everyone can hear. Delivery is whether the singer means it: whether the dynamic arc of the performance follows the emotional arc of the song, whether the quiet lines are actually quiet, whether the big moment is genuinely bigger than what came before.
It shows up as a shape. A take with a flat dynamic profile, every line at the same intensity, sounds correct and uninvolving. A take that builds, holds back, then opens up carries the listener even when it is technically less clean. Given the choice between the two, experienced producers take the second one and fix the pitch.
The fix: there is not one. This is the only dimension of the five with no technical remedy, which is exactly why it deserves the most attention in the room while recording is still possible. You can tune a note, nudge a syllable and de-ess a consonant. You cannot add conviction afterwards.
5. Breath control
Breath is the least glamorous dimension and it silently causes problems attributed to everything else. It shows up in two places.
The first is phrase length. If the singer consistently runs out before the end of a line, the last words lose support, which makes them go flat and lose their consonants. That gets logged as a pitch problem and a diction problem when it is really one breathing problem.
The second is audible breaths in the wrong places. Breath noise is not a defect and removing all of it makes a vocal sound synthetic, but a gasp mid-phrase because the singer had no choice is different from a breath placed deliberately at a comma.
The fix: rework the phrasing so breaths land at natural punctuation, or split the line across two takes. If the singer runs out on every long phrase, the tempo or the key is asking for more than the range comfortably allows.
Putting it together: keep or re-record
The point of separating the five is that they map onto different decisions, and the decision is what you actually needed.
- Pitch problems only: keep the take. Punch in the notes or correct them lightly.
- Timing problems only: keep the take. Nudge the entries that drift.
- Diction problems in solo: re-record the affected lines, it is faster than fighting it in the mix.
- Diction problems only in context: keep the take, fix the arrangement.
- Breath problems: re-record with reworked phrasing, or reconsider the key.
- Delivery problems: re-record, and do it early in the session while the voice is fresh.
Four of the five leave you with a technical task. Only delivery sends you back to the microphone with nothing to adjust but the performance, which is why it is worth listening for first and not last.