Pitch correction basics: graphical vs automatic, and how to keep it natural

Pitch correction is one of the most used and least discussed tools in modern production. It turns up sometimes as an obvious effect and far more often as a quiet fix to a few notes. This guide covers the two main ways of doing it, the handful of settings that decide how it sounds, and how to keep a vocal sounding like a person.

It sits alongside our other mixing basics. If you have not yet set levels and tone, start with EQ basics and compression basics; pitch work is usually done before either.

Two approaches: automatic and graphical

Automatic correction listens to the audio as it plays and pulls each note towards the nearest pitch in a scale you choose. It is fast, works in real time, and is the route to both the transparent "just tidy it up" sound and the famous hard-tuned effect. Antares calls this Auto Mode in Auto-Tune Pro, and says it is the common mode found in all editions of Auto-Tune.

Graphical correction analyses the recording first and draws each note on a grid so you can edit it by hand: move a note up or down, flatten or keep its vibrato, change where it starts. It is slower, but you decide exactly which notes change and by how much. Auto-Tune Pro has a second mode for this, which Antares calls Graph Mode. Celemony built Melodyne around the idea: "In Melodyne, notes are represented by blobs," and you can edit "the pitch, vibrato, volume, sibilants, length, timing and formants of each note." Logic Pro's built-in Flex Pitch works the same way; Apple's user guide says you drag a note vertically to change its pitch and horizontally to move it in time.

A rough rule: automatic for speed, for live use and for the effect; graphical for a lead vocal you care about.

The settings that matter

Key and scale

Automatic correction needs to know which notes are "right". Antares notes that "by default, Auto-Tune will correct to a chromatic scale", meaning every semitone is a valid target. That is safe but loose: a slightly flat note may be pulled to the wrong neighbour. Setting the song's actual key and scale gives the tool fewer, better targets. If you are unsure of the key, find it before you reach for the plug-in.

Retune speed

This is the big one. Antares defines it plainly: "Retune Speed refers to how quickly the plugin adjusts the pitch of an incoming note to correct it." Set it to 0 and, by Antares' own instructions, you get the most pronounced, hard-tuned effect. Slower settings let the natural slide into and out of notes survive. For a natural result, start slow and speed up only until the problem notes sit right.

Humanize and Flex-Tune

Two Auto-Tune controls exist to soften the correction. Antares describes Humanize as something "which slows down retuning during longer sustained notes", and Flex-Tune as something "which makes the retuning effect less aggressive". For the robotic effect Antares tells you to turn both off; for a natural vocal, that suggests leaving them in play. Other tools have their own names for similar ideas, so check your plug-in's manual rather than assuming.

Note-level detail in graphical editors

In a graphical editor you are not setting one speed for the whole take, you are deciding note by note. Apple says each Flex Pitch note has "hotspots" for editing "pitch, vibrato, gain, and other parameters", including one for the pitch drift at the start of a note and one for fine pitch. Melodyne exposes vibrato as its own editable property. Those are the details that make a voice sound human, so treat them carefully.

How to keep it natural

These are habits, not rules from any vendor, and they work in any tool.

1. Fix the take first. Comp the best phrases before you tune. Correction is easier on a good performance and cannot add feeling that was not sung.
2. Correct only what is wrong. In a graphical editor, leave notes that sound fine alone, even if they are not dead centre on the grid.
3. Keep the slides. The scoop into a note and the fall at the end of a phrase are part of a singer's style. Flattening them is the fastest route to a robotic sound.
4. Leave some vibrato. Reducing it a little can tidy a wobbly note; removing it all usually sounds artificial.
5. Listen in context, not solo. A note that looks off on the grid can sound fine against the track, and the reverse.
6. Compare with the original. Bypass the correction regularly. If you cannot tell which version is more convincing, you have probably gone far enough.
7. Watch latency on live use. If you monitor through real-time correction while recording, buffer size matters; our guide to latency and buffer size explains why.

Tuning is one step; once the vocal is in tune, our guide to common bedroom mix mistakes covers what to check next.

Which tool to start with

You may already own one. Logic Pro includes Flex Pitch, per Apple's guide. Auto-Tune and Melodyne are separate products sold in several editions; check each vendor's site for what each edition includes and current prices, as we have not listed features beyond what the pages above say. Check your own DAW's manual before you buy anything.

What we think

Pitch correction has a reputation problem it does not entirely deserve. The hard-tuned sound is a deliberate style, and the transparent kind is closer to editing than to faking. The real risk is not the tool but the habit of tuning everything to the grid because the grid is visible. A graphical editor shows you every note that is a few cents out, and it is tempting to fix them all. Mostly you should not.

Our view: decide what you want before you open the plug-in. If you want the effect, commit to it with a fast retune speed and a correct key. If you want a natural vocal, correct the few notes a listener would actually notice, keep the slides and vibrato that make the singer recognisable, and stop. The best tuned vocal is one nobody thinks about.