Autotune Online for Vocals

The “robotic” sound appears when the pitch is pulled to the note at every moment: the vibrato and scoops are cut off and the voice jumps between notes in steps. That is what the “Hard” sound does — on purpose. With “Natural” each note moves as a whole and its inner movement is kept. If it still sounds artificial, open “Fine-tuning”, lower “Strength” to 60–80 % and raise “Retune speed” to 60–100 ms.

A short take often leaves the key ambiguous: a five-note verse does not contain every degree of the scale. The music contains them all, so the key is detected more reliably from the beat. Besides, many beats are not tuned exactly to 440 Hz — the tool measures the tuning of the backing track in cents and, with “The beat’s own”, puts the voice right into it. A voice that is exact by the standard would sound slightly off over such a beat.

Autotune changes the pitch of the whole sound, so music mixed with the voice gets distorted — the tool warns you when it hears music in the gaps between phrases. If you have the backing track separately, upload the voice as the recording and the beat in the “Backing track” box: the voice is tuned to its key and downloads together with it. If you only have the finished song, split it right here with “Split the voice and the beat” (a Premium feature) or with the “Vocal Remover” tool: the voice gets tuned and the backing track goes back under it unchanged.

Choose the “Robot” sound and add the backing track. The vocoder takes the vowels and consonants from the voice and the pitch from the tuned notes, so the robot sings in the key of the beat. For a monotone “mechanical” voice, press “All notes on one pitch” in “Fine-tuning”, then move single notes on the graph to make your own melody.

This happens when the singer missed by more than a semitone or the key was detected wrong. If the key is wrong, choose the right one in the list or add the backing track: the notes are found again. If one note is wrong, tap it on the graph and move it with the mouse, the ▲ and ▼ buttons or, on a phone, your finger after pressing “Edit notes”. “Leave as sung” keeps that note unchanged.

“Strength” is how much of the distance to the note to cover: 100 % puts the note exactly in tune, 50 % halfway. “Retune speed” is how many milliseconds the voice takes to reach the note: 0 gives an instant jump, 40–80 ms sounds natural. “Vibrato” keeps, weakens or strengthens the voice’s wobble. “Flatten within a note” removes the slow drift on long notes.

Sing to the backing track in headphones rather than speakers, in a quiet room and not too close to the microphone: music that gets into the microphone makes the pitch harder to hear. Echo and heavy reverb get in the way too — add effects after the autotune. The length of the take does not matter: a verse and a whole song are processed the same way.

Yes. Open “No recording? Type a phrase”, type a few words in any language and choose a voice. A synthesized voice says the phrase, and from there it is treated like a recording: the “Hard” and “Robot” sounds turn speech into a sung line, the backing track sets the key, and the notes on the graph can be moved into your own melody. That is how voice tags and short drops for a beat are made.