Autotune Sounds Wrong? Pitch Correction Troubleshooter

← All tools
7 artifacts3 questionsAny pitch pluginFree

Autotune Sounds Wrong? Pitch Correction Troubleshooter

Warble, wrong notes, robotic steps, chipmunk vowels, whistling S sounds. Tell us what you hear and what you want, and get the two most likely causes, the three things to check first, and starting-point settings for Auto-Tune, Waves Tune, Melodyne or your DAW's own tuner.

Diagnose the artifact

Answer all three to see the diagnosis.

Artifact → usual cause → first move

What you hear Usual cause First move
Warble / wobble Note sits between two scale targets; speed too slow for vibrato Set the song key, remove unused notes
Wrong notes Chromatic or wrong scale; nearest target isn't the melody note Thin the scale to the melody notes
Robotic stair-steps Retune speed near 0, Humanize 0, or two tuners stacked Speed 20–60 ms, Humanize up
Chipmunk vowels Formant off, wrong input type, big shift Formant on, fix input type
S and breaths worse Tuner forcing pitch onto noise; no de-esser before it De-ess before the tuner
Fine solo, bad in mix Beat detuned / different key; doubles tuned differently Key-find the instrumental, match settings
Not robotic enough Speed not at 0; Chromatic hides the steps Speed 0, song key, flat delivery out

Control names differ by plugin: Retune Speed / Humanize / Flex-Tune are Auto-Tune terms; Waves Tune uses Speed / Note Transition / Vibrato; Melodyne and Logic Flex Pitch work note by note. The logic is the same.

Why autotune sounds wrong: the five inputs that create artifacts

When autotune sounds wrong it is rarely the plugin being bad. Pitch correction, whether it is Antares Auto-Tune, Waves Tune, Melodyne or the tuner built into your DAW, does one job: detect the pitch of the voice, compare it to a list of allowed notes, and move the pitch toward the nearest one. Every artifact you hear comes from one of five inputs to that decision, and once you know which one is lying to the plugin the fix is usually one control.

Retune speed is how fast the pitch is moved to the target. At 0 ms the move is instant, which is the stair-step T-Pain effect. Slow it down and the plugin follows vibrato and slides, which sounds natural until a note sits halfway between two targets, where the plugin hunts back and forth and you get warble. The honest rule is to set speed for the delivery, not the genre: long held notes with little vibrato can take fast speeds without anyone noticing; slides and runs need Humanize or Flex-Tune to let the transitions through.

Key and scale are the list of targets. Chromatic allows every semitone, which feels safe but is the opposite: a singer 60 cents flat on an E gets pulled to D sharp because that is closer. Setting the key cuts the targets from twelve to seven, and removing notes the melody never uses cuts them further. A detail most tutorials skip: relative major and minor share the same seven notes, so the plugin does not care which you pick. What matters is which notes are in the melody.

Input type sets the detection range (Soprano, Alto/Tenor, Low Male, Instrument). Pick the wrong one and the detector finds the wrong octave on low or breathy phrases, which produces either a wild jump or, with formants off, a chipmunk flip.

Formant is the resonance of the vowel. Shifting pitch without preserving the formant moves the vowel the way a sped-up tape does. Corrections under about 30 cents rarely show it; anything bigger does. Leave it on by default and only turn it off if you want the cartoon sound on purpose.

The source recording is the input the plugin can actually see. Reverb or delay before the tuner, a roomy take, two takes bounced to one file, distortion or a clipped preamp all confuse the detector, and no setting fixes a detector that cannot find the note. The same goes for printed tuning: once a tuned take has been bounced, the artifacts are in the audio and cannot be undone by any plugin after it. Go back to the raw take or re-export dry, then tune again.

Order of operations: Key → Input type → Retune speed → Humanize / Flex-Tune → Formant → De-esser. Changing speed before the key is set is how people spend an hour fixing the wrong thing.

Why hard tuning and sibilance fight each other

S, breaths and the "ch" in "chain" have no pitch; they are noise. A fast retune speed with no Humanize tries to force a pitch onto that noise, and what you hear is a whistle or a thin ringing tail on every S. Resynthesis also makes the consonants more consistent and often louder, so a vocal that was fine raw can suddenly need de-essing. The fix is a de-esser before the tuner so the tuner sees a calmer input, a little Humanize or Flex-Tune so the plugin ignores signal without a clear pitch, and if needed a gentle second de-esser after. Full walkthrough: how to fix sibilance after autotune.

When to use graphical mode instead

Automatic mode makes the same decision on every note. Graphical editing (Melodyne, Auto-Tune Graph Mode, Logic Flex Pitch, FL Newtone) lets you fix the one note that jumps, leave the slides alone and straighten drift on a held note without touching anything else. Go graphical when:

  • one or two notes are wrong in a take that is otherwise good, and raising the speed wrecks the rest;
  • you want a natural result and automatic mode warbles on the vibrato no matter what you set;
  • harmonies or doubles need to land on exactly the same pitch, not just the same scale.

Stay in automatic mode when the tuning is the sound (hard rap and pop hooks) or when most of the take is fine and a sensible speed covers the rest. Many pop vocals use both: Melodyne on the lead for pitch, then Auto-Tune in automatic mode on top for the polish. See how to tune vocals in Melodyne and the retune speed guide.

FAQ

Why does autotune warble between notes?

The sung pitch is sitting between two allowed targets and the plugin keeps changing its mind, usually because the scale is Chromatic or wrong, or because the retune speed is too slow for a note with vibrato. Set the song key, remove notes not in the melody, then shorten the speed only on the phrase that warbles.

Why does Auto-Tune jump to the wrong note?

Because a different note in the scale was closer to what was sung. Big intervals, falsetto flips and ad-libs land far enough off that the nearest target is not the one the singer meant. Thin the scale to the notes the melody uses, and fix the one stubborn line in graphical mode.

Should I leave the scale on Chromatic?

Only when you genuinely do not know the key or the melody is very chromatic. For most songs, Chromatic gives the plugin twelve targets instead of seven, which produces more wrong notes and more warble, not fewer. Use the song key finder on the instrumental and treat the answer as a hint to confirm by ear.

What retune speed sounds natural?

Roughly 20–60 ms with Humanize somewhere between 30 and 70, adjusted per phrase. Pop and rap polish usually sits around 5–20 ms. The classic hard effect is 0 ms with Humanize and Flex-Tune at 0. Numbers vary with the singer, so treat these as starting points.

Can I fix tuning that was already printed or bounced?

No. Once the tuned audio is rendered the artifacts are part of the file, and running a second tuner on top makes it worse. Go back to the dry take, or ask whoever recorded it to re-export without tuning, then tune again.

Why does my voice sound like a chipmunk after tuning?

Formant correction is off, or the input type is wrong so the plugin corrected by an octave. Turn Formant on, set the input type to your actual voice range, and keep transpositions within about three semitones.

Will pitch correction make me sound like a specific artist?

No. Pitch correction changes pitch, not timbre, and no setting or plugin turns one voice into another. What you can copy is the approach: the key, the speed, the plugin order and the processing around it, which is what the hard-tuned chains in the Vocal Chain Bible document.

Streams follow sound

Know the problem. Now copy the chain that solves it.

The Vocal Chain Bible is a 187-page PDF with 88 real vocal chains from hit records, as the engineers reported them, with exact mic, preamp, plugins and settings, each in a paid and a free-plugin version. It includes the Autotune Hip-Hop Vocal Chain Guide and the hard-tuned rap chains, so once you know why the tuning sounds wrong you can see where the tuner sits in a chain that works.

187 pages · 88 chains · Try one free chain first
Copied

Free download

Get the free Grammy Sauce PDF

The plugins and settings 100+ Grammy-winning engineers actually use — collected from 5 years of interviews, in one free PDF.

Want the full pro setup? Get the Vocal Chain Bible — 88 real chains →

Contact form