Vocal editing: how to comp takes and fix timing without killing the performance

Most vocals you love on records are not one take. They are a comp: the best line from take three, the best word from take five, stitched so cleanly that you hear a single performance. Comping, and the light timing work that follows it, is the least glamorous part of vocal production and one of the most important. A great comp on a plain chain beats a weak take through an expensive one.

This guide covers the stage between recording vocals at home and building your vocal chain. The examples use Ableton Live because its manual is clear on the mechanics, but every major DAW has the same tools under different names. Logic Pro, for instance, does this with take folders, which Apple documents in its comping overview.

Step 1: Record for the comp, not for the perfect take

Comping starts before you edit. Record a section, not the whole song in one pass, and loop it three to five times. That gives you options for every line without wearing out the singer.

In Live, you do not need to set anything up. The Ableton manual says "take lanes are automatically created when recording new clips in the Arrangement", and recording over a loop adds a new lane for each pass. Other DAWs stack takes in a similar way.

Some habits that make the edit easier:

• Keep the same mic position and input gain across takes, so lines from different passes match in tone and level.
• Name or colour takes as you go ("verse 1, warm-up", "verse 1, keeper").
• Note the bars where the singer felt something special. Your memory of the session is data.
• Stop at five or so takes per section. Past that, you are usually choosing between tired versions of the same thing.

Step 2: Listen first, then mark

Do not start cutting on the first listen. Play each take of a section through once with your eyes off the screen. Then go line by line and mark the best version of each phrase. A simple rating works: a tick for usable, a star for special.

Choose for emotion first, then pitch, then timing. Pitch and timing can be nudged later; a flat delivery cannot be fixed. If one take has the right feeling but a sour note, it often still wins.

In Live you can audition a single lane with its speaker button, then select the part you want and, as the manual puts it, "selected material in take lanes can be copied to the main lane by pressing the Enter key." Work phrase by phrase until the main lane holds a full verse.

Step 3: Make the edits invisible

A comp is only as good as its joins. Three rules cover most problems:

1. Cut on breaths or consonants. The gap before a word, or a hard consonant such as "t", "k" or "p", hides an edit well. Cutting in the middle of a held vowel almost always shows.
2. Use short crossfades on every join. A few milliseconds removes clicks. Lengthen the fade if you still hear a bump.
3. Keep the breath from the same take as the line that follows it. A breath tells the listener how the next phrase will sound; a mismatched one feels wrong even when nobody can say why.

Then listen to the whole comp in context, with the beat, at normal volume. If any line sounds like it was sung in a different room or a different mood, swap it, even if it was technically the best.

Step 4: Clean up, but do not sterilise

With the comp settled, tidy the track:

• Remove noise between phrases: chair creaks, headphone bleed, page turns. Cut or fade rather than gate, so you control each one.
• Turn breaths down rather than deleting them. A vocal with no breaths sounds pasted together. Clip gain or a quick volume dip works, or use the techniques in our automation guide.
• Even out obvious level jumps between words with clip gain before any compressor sees them. The compressor will then work on tone, not on fixing your comp.

Step 5: Fix timing only where it matters

Vocals sit naturally around the beat, not on it. Do not quantise a lead vocal to the grid. Instead, listen for three specific problems:

• Phrase starts that are clearly late or early, especially on the first word after a gap.
• Consonants that clash with the drums, such as a "t" landing just before a snare.
• Doubles and harmonies that do not line up with the lead, which smear the sound and make the vocal feel messy.

For a whole phrase, simply slide the clip. For a single word, stretch it. In Live this is done with warp markers; the manual notes that "transients are usually good places to put Warp Markers" and that "you can manually add any number of markers as needed"; you then drag a marker to move the audio around it. Put a marker before and after the word you want to move, then move the one at the start of the word. Small moves are the goal. If the stretch is audible, undo it and cut the word into its own clip instead.

Doubles are the exception to "leave it loose". Tighten them to the lead at the starts and, above all, the ends of words, where a long "s" from two voices sounds like a hiss. A common fix is to trim the sibilants from the double and let the lead carry them.

Step 6: Then, and only then, tune

Pitch correction goes after comping and timing, because moving audio around changes what the tuning plugin sees. If you tune first, you redo the work. Our pitch correction basics guide covers that step.

A quick checklist

1. Record sections in loops of three to five takes, same gain and mic position.
2. Listen through once before cutting.
3. Mark the best phrase for feeling first.
4. Cut on breaths or consonants, crossfade every join.
5. Level with clip gain, keep breaths quieter but present.
6. Move phrases and words only where you can hear a problem.
7. Tighten doubles to the lead, especially word endings.
8. Tune last.

What we think

The common advice is that comping is about finding the best take for each line. We think that framing quietly trains people to build vocals that are correct and lifeless. The real job is to keep one consistent character across the song: the same singer, in the same mood, telling one story. A line that is slightly rougher but matches the takes around it is usually a better choice than the cleanest line from a pass where the singer had moved on emotionally.

That is also why we would rather you comp from fewer takes. Five passes recorded close together share an energy; twenty passes across an evening do not, and the comp ends up as a collage of different people. Editing tools keep getting faster; use them to save clicks, not to make the choice. The decision about which version of a line tells the truth is the singer's and the producer's, and it is the part of the record listeners actually feel.