Pitch correction tools should be set to correct a vocal to within about 10 cents of the target note rather than to zero, because a perfectly tuned vocal sounds synthetic and lifeless to the human ear. That single number, 10 cents, is the difference between tuning nobody notices and tuning everybody notices. Retune speed is the knob that controls it, and almost every robotic-sounding home recording has that knob set far too fast.
The other thing worth saying up front: pitch correction has a hard ceiling that no other studio process has. You can compress a wildly uneven take into something even. You cannot tune an out-of-tune take into something in tune without audible cost, because every cent of correction generates artifacts. More on why that matters later.
Quick answer box
- Correct to 10 cents, not 0 cents. Perfect is worse.
- Retune speed 20 to 40 ms is transparent for most pop and rock leads.
- Turn formant preservation on, always. It is why voices go chipmunky.
- Tune background stacks harder than leads. Layers beat against each other.
- Under 5 cents of drift is inaudible. Do not correct it.
Some links on this page are affiliate links. If you buy through them, we may earn a commission at no additional cost to you.
The 10-Cent Rule
A cent is one hundredth of a semitone. Most singers have never had this explained, which is why "how much correction is too much" feels like a mystery instead of a number.
Here is the perception scale.
| Deviation from the target note | What a listener actually perceives |
|---|---|
| 0 to 5 cents | Inaudible. Leave it completely alone. |
| 5 to 15 cents | Reads as human warmth on sustained notes |
| 15 to 30 cents | Audibly off. This is the range worth correcting. |
| 30 to 50 cents | Clearly flat or sharp to any listener |
| Over 50 cents | Approaching a different note entirely |
The commercial vocals you are comparing yourself to are not tuned to zero. They typically sit somewhere in that 5 to 15 cent band, with the drift preserved deliberately. Human hearing reads tiny pitch variation as expression; remove it, and the voice stops sounding like a person and starts sounding like a sample.
So the goal is not accuracy. The goal is landing inside a window. Correct the 15 to 30 cent problems, leave the 5 to 15 cent life alone, and never touch anything under 5 cents.
Automatic vs Manual: The Distinction That Actually Matters
Every list of pitch correction software ranks products by price. That is the wrong axis. There are two fundamentally different kinds of tools here, and picking the wrong category is a bigger mistake than picking the wrong brand.
| Automatic (real-time) | Manual (offline, note-based) | |
|---|---|---|
| Examples | Auto-Tune, Waves Tune Real-Time, Graillon, GSnap, MAutoPitch | Melodyne, Logic's Flex Pitch, Cubase VariAudio |
| How it works | Snaps incoming pitch to a scale as it plays | Detects individual notes; you drag each one |
| Time per song | Seconds | 10 to 40 minutes |
| Transparency | Entirely dependent on retune speed | Very high when used carefully |
| Handles a genuinely wrong note | Poorly | Well |
| Preserves vibrato | Only at slow retune speeds | Yes, and vibrato is editable separately |
| Produces the recognizable "effect" | Yes, that is its origin | Not really |
| Best used on | Background stacks, live use, deliberate stylization | Lead vocals, releases, individual problem notes |
The practical workflow most engineers use: manual note editing on the lead vocal, automatic correction on the background stacks. Manual because the lead carries the song and deserves the time; automatic because nobody is listening closely enough to a fifth-layer harmony to catch artifacts.
If you are choosing a DAW partly on what pitch editing it includes, several free options ship with genuinely capable note editors. That comparison lives in the best free recording software for singers.
Retune Speed by Material
This is the one setting that decides whether your correction is invisible or infamous. Start here, then adjust by ear.
| Retune speed | Result | Use it on |
|---|---|---|
| 0 to 5 ms | Hard instant snap; the recognizable effect | Trap, hyperpop, deliberate stylization |
| 5 to 15 ms | Tight, modern, slightly synthetic sheen | Modern pop leads and hooks |
| 15 to 30 ms | Mostly transparent, controls drift well | Most pop and R&B leads |
| 30 to 60 ms | Transparent; vibrato survives intact | Ballads, rock, singer-songwriter |
| 60 to 120 ms | Very gentle; catches only long drifts | Jazz, folk, acoustic |
The mistake I see constantly is a singer-songwriter recording their acoustic ballad with retune speed at the plugin default, which on many tools is fast. Their vibrato gets flattened into a straight line, the note entries lose their natural scoop, and they cannot work out why the vocal sounds sterile. It is not the mic. It is a number they never touched.
Rule of thumb: the slower and more expressive the song, the slower your retune speed should be.
What Pitch Correction Cannot Fix
Every correction has a cost. Here is what breaks first, by problem type. This is my own scoring based on where artifacts start becoming audible.
| The problem | How much correction stays transparent | What breaks first |
|---|---|---|
| Sustained note slightly flat | A lot, up to roughly 50 cents | Almost nothing. This is the easy case. |
| Note entry scooped from below | Moderate | The scoop flattens; entries sound abrupt and clipped |
| Vibrato drifting sharp | Small | Vibrato squashes into a straight line |
| Wrong note by a semitone or more | Very small | Formants shift, producing chipmunk or gargle artifacts |
| Breathy or whispered passages | Very small | The algorithm cannot find a pitch and warbles |
| Two voices bleeding onto one track | None at all | Monophonic detection fails completely |
| Timing that rushes or drags | Not applicable | Nothing. This is a time-editing problem, not a pitch one. |
Two rows there deserve extra attention.
Breathy passages are the hidden trap. Pitch detection needs a clear fundamental frequency to lock onto, and a whispered or heavily breathy phrase does not have one. The plugin guesses, and the guessing is what you hear as warbling. If you plan to tune a track, sing it with more tone than you think you need.
Bleed between two singers cannot be tuned by monophonic tools at all, which is worth knowing before you record two people on one mic to save time.
The Formant Trap
This is the single most common cause of "my tuned vocal sounds like a cartoon."
Formants are the resonant peaks created by the shape of your throat, mouth, and nasal cavity. They stay roughly in the same place regardless of what pitch you sing, which is a large part of why your voice is recognizably yours. A naive pitch shift drags the formants along with the note, which shrinks or enlarges the apparent size of the singer's head. Hence, chipmunk on the way up and gargling monster on the way down.
Every modern tool has formant preservation. Turn it on and leave it on.
Even then, it has limits. Formant preservation stays convincing for roughly two to three semitones of shift and degrades noticeably beyond that. If you need to move a note further than that, re-sing the line. It will take less time than fighting the artifacts.
The signal processing underneath all of this comes out of decades of academic research; Stanford's Center for Computer Research in Music and Acoustics publishes on exactly these areas, including psychoacoustics and the signal processing behind pitch detection, if you want to understand the mechanism rather than just the settings.
Tuning a Vocal Stack Differently
Leads and stacks get opposite treatment, and this catches almost everyone out.
Leads: tune lightly. Small drift reads as human on a solo voice. Aggressive correction on an exposed lead is the most audible mistake in the whole process.
Stacks: tune tightly. When six layers each sit 12 cents off in different directions, they beat against each other and the whole stack sounds sour rather than warm. What reads as humanity on one voice reads as out of tune across six. Push retune speed faster and correction harder on background layers, and nobody will ever notice.
The same inversion applies to reverb and compression on stacks, which is covered in layering vocals and double-tracking.
Order of operations: tune before you compress and EQ. A compressor reacting to a pitch-corrected signal behaves differently than one reacting to the raw take, and re-tuning after you have built a chain means redoing the chain. The full processing order is in EQ and compression on vocals.
The Correction Audit
Here is the diagnostic, and it is free.
Open your most recent song. Look at your tuning session and count two things: how many notes you corrected, and the largest single correction you made.
- Fewer than 10 notes, none over 20 cents. You are using pitch correction as polish. This is the healthy zone. Carry on.
- 10 to 30 notes, several in the 20 to 40 cent range. Borderline. Artifacts are creeping in even if you have not consciously noticed them yet. Worth addressing.
- More than 30 notes, or any correction over 50 cents. You are not correcting a melody; you are rebuilding one. No plugin setting fixes this.
That third result is the one worth being honest about, because of the ceiling I mentioned at the top. Every other stage of production has a workaround. A weak room can be treated. An uneven performance can be compressed. A dull recording can be EQ'd. Pitch is the exception: the correction itself is the thing that sounds amateur, and correcting harder makes it worse rather than better.
Which means pitch is the one area where training genuinely outperforms software, and it is not close.
Start free. Singing Carrots has a browser-based pitch monitor that shows your deviation in real time while you sing. Run through your problem phrase and watch where you drift. Costs nothing, takes ten minutes, and tells you whether your issue is one specific interval or general instability.
If you cannot hear the drift, that is the root cause, and it is fixable. EarMaster drills pitch and interval recognition in a structured order. This is the highest-value purchase on this page by some distance, because a singer who can hear a 15-cent drift stops producing takes that need heavy correction in the first place.
If you can hear it but cannot hold it, the cause is usually breath support or placement rather than perception:
- 30 Day Singer for daily structured practice; pitch stability is built through consistent reps more than through insight.
- Singorama if you want breathing, pitch, and tone handled in one course.
- Roger Love's program if you drift specifically on recorded takes while sounding fine live, which usually points to placement changing under headphones.
- Robert Lunte / The Vocalist Studio if you go sharp or flat on belted and distorted notes at the top of your range.
Worth saying plainly: this recommendation costs me gear affiliate revenue, and it is still the right advice. Software is the cheaper purchase and the worse investment here.
Common Mistakes
Leaving retune speed at the default. The default is fast on many plugins because fast demos work well. It is wrong for most music.
Tuning to zero. Covered above, and it is the reason most "professionally tuned" home vocals sound worse than untuned ones.
Setting the wrong key or scale. An automatic tool set to the wrong scale will confidently snap your notes to the wrong pitches. Check the key before anything else.
Tuning a bad comp. Pick your takes first. There is no point tuning a phrase you are going to replace, and a better take usually needs less correction anyway. The comping process is in how to record vocals at home.
Correcting breaths and consonants. Automatic tools will happily try to tune a breath. Bypass those regions or use manual editing.
Tuning before comping and after processing. Both are backwards. Comp, then tune, then process.
Using correction to hide a range problem. If a song sits above your comfortable range, every note in the chorus will need correction. Transposing the song down two semitones is free and sounds better than tuning forty notes.
Frequently Asked Questions
What are the best pitch correction tools for beginners?
Start with whatever ships in your DAW, since most modern DAWs include a capable note editor. Free plugins like GSnap, MAutoPitch, and the free version of Graillon cover automatic correction well enough that buying a paid tool should wait until you know exactly which feature you are missing.
How much pitch correction is too much?
Correcting more than about 30 notes in a song, or any single note by more than 50 cents, is past the point where the correction itself becomes the audible problem. Aim to land within 10 cents of the target rather than at zero.
Why does my tuned vocal sound robotic?
Almost always, retune speed is set too fast. Try 30 ms and listen again. The second most common cause is correcting to zero deviation, which strips out the micro-variation that makes a voice sound human.
Do professional singers use pitch correction?
Widely, yes, but usually lightly, and often through manual note editing rather than automatic correction. The tuning you do not notice on commercial records is the norm; the tuning you do notice is a deliberate stylistic choice.
Can pitch correction fix a bad vocal take?
No. It fixes small deviations in an otherwise good take. Wrong notes, breathy passages, and unstable pitch all produce audible artifacts when corrected heavily, which is why re-recording is usually faster than fighting the plugin.
Should I tune before or after compression?
Before. Tune the comped raw vocal, then build your processing chain on top of the corrected signal.
Your Next Step
Pitch correction tools work best when they are barely working at all. Set retune speed to match your material, keep formant preservation on, correct into a 10-cent window rather than to zero, and treat heavy correction as a signal to re-record rather than a setting to push harder.
Run the free pitch monitor on the phrase you always struggle with. Five minutes, no cost, and it will tell you whether you have a plugin problem or a practice problem.
Then continue through the recording and music production guides for singers, or work on the capture side first with how to make your voice sound better in recordings. If budget is the constraint, recording vocals under $200 covers the free and low-cost path in full.
