Autotune online

Autotune a vocal in three steps

1. Choose a vocal recording (MP3, WAV, M4A, a voice memo or a video), or try the sample: a synthesized singer with every note deliberately 20–40 cents off.
2. Pick Chromatic to snap to the nearest semitone, or choose the song’s key and scale so notes only land on notes of that scale. The page suggests a key from the notes you sang. Set the retune speed: 0 ms for the hard effect, 20–60 ms for clear correction, 100 ms or more to keep vibrato.
3. Press Apply autotune, compare the pitch chart and the preview with the original, and download the tuned vocal as MP3 or WAV.

How this autotune works

First the page tracks the pitch of your voice every 5 milliseconds with the McLeod Pitch Method, the same engine as our pitch detector. For each moment it picks the target note: the nearest semitone, or the nearest note of the key and scale you chose, decided from the pitch averaged over about a tenth of a second so vibrato does not flip it between two notes. The retune speed smooths the correction over time. Then the voice is resynthesized with TD-PSOLA (pitch-synchronous overlap-add): the recording is cut into small overlapping pieces, one per vibration of your vocal cords, and the pieces are laid back down slightly closer together to raise the pitch or further apart to lower it. Because each piece keeps its own shape, the resonances of your mouth and throat stay where they were, so a corrected voice does not turn into a chipmunk.

Retune speed: natural or T-Pain

Retune speed is how quickly the pitch is pulled to the target. At 0 ms every note snaps instantly and stays perfectly flat, and jumps between notes become steps: that is the hard, robotic sound heard on records by T-Pain and on Cher’s “Believe”. At 20–60 ms, notes land in tune but slides between them still sound sung. From about 100 ms, vibrato and expressive scoops pass through and only the centre of each held note is moved, which is closer to quiet studio correction. Choosing the right key matters more than speed: in the wrong key, notes are pulled to the wrong place. If you are not sure, use Chromatic, which never moves a note by more than half a semitone.

What we tested, and the limits

On our own synthesized vocal, with every note 18–38 cents off, the median distance from the nearest note dropped from 28 cents to about 1 cent at retune speed 0, measured back with the same pitch detector, with the vowel resonances unchanged. Real voices are harder. Breathy, raspy or whispered singing has no clear pitch, so those moments are left as they are. Fast runs and ornaments may be flattened at hard settings. Reverb, double-tracked vocals, harmonies and backing music confuse the pitch tracker and can cause warbles, which is why a dry solo vocal matters. This is not a substitute for a studio plugin with manual note editing, and real singers on many microphones have not been tested yet.

Privacy and trademarks

The recording is decoded and processed inside your browser tab; nothing is uploaded and there is no account. Auto-Tune is a registered trademark of Antares Audio Technologies. AudioClover is not affiliated with Antares; we use “autotune” in its everyday sense of automatic pitch correction.