01
Natural correction
Strength 35–60% · Retune 30–50 ms · Humanize 10–20 cents. For a singer who's mostly in tune and needs the few sour notes cleaned up without anyone noticing. Keep vibrato high.
Autotune your vocals
Upload a song, pick one of yours, or sing it right here — then hit render.
Drop your track here
MP3 / WAV / FLAC / M4A · max 50 MB · 10 min
Sing straight into your mic — up to 7 minutes, saved as a clean WAV.
My audio is a dry vocal
30 credits · about 2–3 min
We run AI vocal separation first, then tune only the voice and mix it back over the original instrumental.
Tuning an already-separated vocal sounds cleaner than tuning a full mix — and costs less. Separate it first
Strength70%
Retune speed25 ms
Results
Compare against the original, then download.
Free Online Autotune — for whole songs, not just acapellas
Drop in a finished track and we separate the vocal, tune it, and mix it back. Transparent pitch correction that still sounds human, or the hard T-Pain effect — your call. No plugins, no watermark, WAV export.
How it works
Three steps, no plugins
Everything runs on our servers. Nothing to install, and it works the same on a phone.
1Step 1
Upload a song or a dry vocal
Most people only have the finished track, so that's the default. We run AI vocal separation, tune only the voice, and mix it straight back over the original instrumental. If you already have an isolated acapella, pick that instead — it's faster and cheaper.
2Step 2
Pick a preset, or set it yourself
Natural is the default: audibly in tune, still recognisably you. One click gets you the hard T-Pain or Travis effect. Every slider underneath stays available — strength, retune speed, vibrato, formant, and a safety clamp.
3Step 3
Compare, then download
A/B between the original and the tuned version with no gap and no jump in playback position, and read the pitch curve to see exactly what moved. Export MP3 320 on any plan, or WAV 16 and WAV 24 with a Basic plan or above — no watermark either way.
The controls
What each setting actually does
Most online tuners give you two sliders and no explanation. Here's the whole panel, in plain terms.
Strength
How far a note gets pulled toward the correct pitch. At 100% every note lands exactly on the grid, which is what the hard effect needs. Below about 60% most listeners can't hear that anything happened at all, which is why so many free tools feel broken — they ship a default that does nothing. We start at 70%: clearly working, still sung.
Retune speed
How quickly the pitch snaps once a note begins, in milliseconds. Zero is instant and gives you the stepped, robotic T-Pain sound. Twenty to thirty milliseconds is what engineers use for correction you aren't supposed to notice. Go past sixty and fast rap phrasing slips through uncorrected.
Humanize
Fast retune is required for short notes but ruins long ones — a held note locked dead flat for three seconds sounds synthetic. Humanize slows the correction down in proportion to how long a note is held, so the fast phrases stay tight and the sustained notes keep drifting the way a real voice does. Almost no online tuner exposes this.
Tolerance
A dead zone, measured in cents. Anything already within this distance of correct is left completely alone — not corrected slightly, not touched at all. This is what keeps deliberate expression alive. Blues and jazz vocals bend pitch on purpose; raise tolerance to 30–50 cents and those bends survive while the genuinely wrong notes still get fixed.
Vibrato keep
Vibrato is a fast wobble of roughly 5 Hz sitting on top of the note. Correction that doesn't separate the two either flattens the wobble or fails to fix the note. We split the pitch into a slow baseline and the vibrato riding on it, correct only the baseline, then add your vibrato back at whatever proportion you choose. This is the single control that decides whether 'natural' actually sounds natural.
Formant shift
Pitch is how high a note is; formants are the resonances that make it sound like your throat. Naive pitch shifting drags them together, which is why cheap tools give you the chipmunk effect. Ours keeps formants locked by default, and exposes them as a separate slider — negative for a bigger, deeper voice, positive for a smaller, brighter one, at any pitch. Competitors either hide this behind a paywall or don't offer it at all.
Starting points
Three settings that work
Presets get you close. If you want to dial it in yourself, start here.
02
Modern rap and melodic trap
Strength 70–100% · Retune 5–20 ms · Humanize 5–12 cents. Tight and obviously processed, but still short of full robot. Drop tolerance to zero so nothing slips past.
03
Stacks and doubles
Strength 60–90% · Retune 15–30 ms · Humanize 6–14 cents. Backing layers need to sit tighter than the lead or they smear. Tune the doubles harder than the main vocal.
Pricing
What's free and what costs credits
No trial countdown, no surprise paywall halfway through.
Free
Listening to the before/after examples. Uploading a track or recording one here. Changing every setting, including the advanced ones. Reading the pitch curve and the spectrograms on any result you've rendered.
15 credits
Rendering a dry vocal end to end. No separation is needed, so you only pay for the tuning — the same 15 whichever preset you pick. The export format doesn't change the price; WAV needs a Basic plan or above.
30 credits
Rendering a full song end to end: AI vocal separation, tuning, and mixing back over the instrumental. That's 15 for the separation plus 15 for the tuning — the same rate you'd pay for each on its own.
FAQ
Questions people actually ask
You get monthly credits free when you sign up, so your first tracks cost nothing out of pocket. Credits are only spent on the render itself: 15 for a dry vocal, 30 for a full song — that's 15 for the AI vocal separation plus 15 for the tuning, the same rate each would cost on its own. Uploading, recording, changing every setting and reading the pitch curve are all free, and if a render fails the credits go straight back.
No. Nothing is added to the audio, at any tier — the MP3 320 comes out as clean as the WAV. WAV 16-bit and 24-bit export come with a Basic plan or above.
A whole song works, and it's the default here. We run AI vocal separation first, tune only the isolated voice, then mix it back over the untouched instrumental. Almost every other online tuner requires you to bring your own acapella — which is the real reason people give up on them. If you do already have a dry vocal, pick that mode: it's faster and half the price.
Leave Key on Auto-detect. We read the key from the full mix, where the chords are, which is meaningfully more accurate than reading it from an isolated vocal. When the detection isn't confident we fall back to chromatic — every semitone allowed — and tell you so, rather than guessing and dragging your vocal onto the wrong notes. You can always override it.
Auto-Tune is a plugin you buy and run inside a DAW, with note-by-note graphical editing. This runs in a browser, takes a finished song rather than an isolated track, and gives you the parameters that matter for a quick result rather than a full editing surface. If you're mixing an album professionally, buy the plugin. If you want a demo in tune in two minutes, this is faster.
Two causes. First, pitch detection occasionally misreads a note by an octave — common on deep male voices and breathy singing — and the tuner obediently drags the whole phrase there. We clamp the maximum shift and repair octave jumps before anything is applied. Second, unvoiced consonants like s and t have no pitch at all; feeding them through a pitch shifter produces a metallic whistle. We detect those and pass the original waveform through untouched.
Not unless you ask it to. We separate the slow pitch baseline from the vibrato riding on top, correct only the baseline, and add your vibrato back at whatever proportion Vibrato keep is set to. The hard presets set it to zero on purpose — that's part of the effect.
Yes. All the processing happens on our servers, so there's nothing to install and no difference between phone and desktop. Nothing depends on your device's audio hardware.
It's processed on our servers, working files are deleted after processing, and nothing is used to train models. Your results stay in your own history until you delete them.
Yes — it's one click. The robotic quality doesn't come from setting retune speed to zero, which is what most tutorials tell you; it comes from holding each note at a single fixed pitch for its whole duration. Our T-Pain, Travis Scott and Cher presets all do that properly, which is why they sound like the records.
Related tools
The rest of the vocal chain
Vocal Remover
Pull the vocal or the instrumental out of any song.
Stem Splitter
Split a track into vocals, drums, bass and more.
AI Mastering
Master your finished mix against a reference track.
LRC Generator
Get word-level synced lyrics from any audio.
MP3 to MIDI
Turn a melody into editable MIDI notes.
Lyric Video Maker
Turn a finished song into a lyric video.
MP3 Tag Editor
Fix the title, artist and cover art on the file you just tuned.