Auto-Tune — Pitch Correction, Download & A/B Compare in Your Browser (No Upload)
Free online auto-tune — vocal pitch correction in your browser, no upload. Correct a dry vocal to any key and scale with formant-preserving pitch correction. Record or upload in the header Controller, press Apply, compare on the pitch roadmap, then export with Download Auto-Tune (MP3, WAV, OGG, or M4A), Save Auto-Tuned as Take into slot A–D, or from the Controller download menu.
Correction runs in the tab, not on a server
Pitch detection and TD-PSOLA correction both run locally with WebAssembly. The audio stays in your browser tab and is never sent anywhere. A slow connection does not add queue time — the only wait is your own device analyzing a longer take, which the progress bar follows.
Hear what your take sounds like in tune — without a DAW
This won't replace Auto-Tune Pro or Melodyne on a complex master. It uses TD-PSOLA so your vocal color stays closer to your voice while pitch moves toward the scale. Use it to hear a phrase corrected to a key, spot flat notes, or try Gentle and Natural presets when you want a light pass — not a robotic snap.
Play the original and the corrected version back to back
After Apply, use Load Original and Load Auto-Tuned to switch versions in the header player at the same timestamp. The correction stays on this page until you press Save Auto-Tuned as Take, choose a slot, and name it. The pitch roadmap shows the original trace in gray and the corrected trace in pink. Click the chart to move the playhead.
Download the corrected file without leaving the browser
After Apply, use Download Auto-Tune beside Re-apply: pick Format (MP3, WAV, OGG, M4A), then download. Filenames use the take’s recording time, e.g. autotune_2025-10-03_14-30-00.mp3. Or press Save Auto-Tuned as Take, pick slot A–D, and export that slot from the Controller download menu — same local encode, no upload.
How to use it
Record a take using the Controller strip in the header, or upload an existing audio file. Once a recording is loaded, the controls below become active. Set your key and scale — or pick a preset — then press Apply Auto-Tune. The corrected file is generated in your browser. Nothing is uploaded. A dry vocal is the right input; a full mix gets pitch-shifted as a whole.
What each control actually does to your voice
- Key — Tell the tool what key the song is in. If you're singing a C major melody, pick C. Getting this wrong means your notes will be pulled toward the wrong pitches — you'll hear it immediately. If you're not sure of the key, set Scale to Chromatic: it snaps each note to the nearest semitone without forcing it into a specific key.
- Scale — Defines which notes are allowed after correction. Major is the bright, "happy" sound most Western pop uses. Minor (natural or harmonic) is darker and more dramatic. Pentatonic is the five-note scale behind most rock, R&B, and soul leads — it's hard to go wrong with it. Blues adds the blue note for that characteristic bent feeling. Chromatic doesn't restrict anything, just moves each pitch to the closest half step.
- Strength (0–100%) — How far the center of each note moves toward the scale. At 100% the center sits on the target note. Vibrato and scoops are not removed here; Retune Speed does that. Around 60–75% the center moves most of the way. Below 40% it barely moves. The first visit starts at 75%. A saved setting in this browser is kept.
- Retune Speed (0–100%) — How fast that center arrives, and how much vibrato stays. 0% snaps at once and flattens vibrato — the robotic sound. 100% eases in and keeps vibrato and short ornaments. The slider is labeled Snap on the left and Vibrato on the right. 50–70% is the natural range.
- Tolerance (0–100¢) — A dead zone around each target note. Notes that land within this distance (in cents — hundredths of a semitone) are left completely alone. Only pitches further away get corrected. This is useful for blues and jazz where slight pitch bends are intentional: raise Tolerance to 30–50¢ and those deliberate inflections won't get snapped flat.
Before you hit Apply
- You need an active take from the header Controller — record, upload, or select slot A–D in Takes and backing. Apply stays grayed out until that take is loaded and playback is paused.
- If you don't know the key, Chromatic is a safe default. It corrects pitch without forcing a specific tonal center.
- Presets — Gentle, Natural, Moderate (Default), Tight, and Hard set Strength, Retune Speed, and Tolerance together. Gentle is a light pass: Strength 30%, Retune Speed 95%, Tolerance 20¢, so only a little of the drift moves and vibrato stays. Moderate (Default) is 75% / 55% / 25¢, the same start a first visit uses. Tight is 95% / 25% / 10¢: notes land on pitch for a polished pop sound, without the stepped robotic effect of Hard. Sliders still override a preset. If this browser already saved settings, those values load instead.
- A good starting point when nothing is saved yet: Strength 75%, Retune Speed 55%, Tolerance 25¢.
After processing
- Intonation readout — average distance to the nearest scale tone, measured on the original and again on the corrected audio.
- Pitch roadmap — gray is the original pitch, pink is the pitch measured on the corrected file. Click the chart to move the playhead. Gaps are unvoiced sections (consonants, breath).
- Load Original / Load Auto-Tuned — pushes either version into the Controller player so you can compare them at the same timestamp.
- Re-apply Auto-Tune — change Key, Scale, sliders, or preset and run again. Always starts from the original audio; corrections do not stack.
- Download Auto-Tune — appears in the action row after Apply: Download Auto-Tune button, then Format (MP3 → WAV → OGG → M4A). Saves the corrected file locally; filename uses the take recording timestamp (
autotune_YYYY-MM-DD_HH-MM-SS.ext). - Save Auto-Tuned as Take — after Apply, stores the corrected audio in a slot A–D with a name you choose. It does not replace the dry recording in another slot unless you overwrite that slot on purpose.
- Controller export — after Save Auto-Tuned as Take, open Download in the header, pick that slot, and choose MP3, WAV, OGG, or M4A (WebM is for mic captures, not re-encoded corrections).
Who ends up using this page
- Singers checking a practice take and seeing which notes kept landing flat or sharp before correction.
- Songwriters cleaning up a rough demo's pitch before sending it to a collaborator.
- Voice teachers showing a before/after of intonation work during a lesson.
- Anyone curious what their own voice sounds like corrected, without buying a plugin.
Reading the roadmap instead of just the result
- Record a phrase dry, apply a Gentle preset, and compare the pitch roadmap to see exactly which notes the engine had to move the most — those are the ones worth drilling without correction.
- Use Hard as a diagnostic, not a destination: a phrase that still sounds choppy under heavy correction usually has a timing or breath issue correction cannot fix.
- Save a corrected take as a reference, then try to match its pitch center by ear on the next live attempt.
Recording a dry take versus uploading one
Either path works, because Apply just needs a dry, monophonic vocal in the Controller — it does not care whether that take arrived by recording or by upload. Record live when you want to try a phrase, hear it corrected, and immediately try again with a different Strength or Scale. Upload when the take already exists: a vocal stem from a session, a demo you tracked elsewhere, or a dry recording a singer sent you. The one thing both paths need in common is that the file stays monophonic vocal only — a file with backing already mixed in gets pitch-shifted as a whole, instruments included.
What Apply does to the file
Apply walks the take, estimates the fundamental with YIN, and builds a target pitch on the key and scale. The resynthesis is TD-PSOLA: it marks pitch periods in the waveform and overlaps those periods closer together to go up, or further apart to go down. Because that edit is on the time-domain periods, the vocal-tract resonances move less than they would if the whole spectrum were simply resampled. Large jumps and noisy takes still sound rough. Polyphonic audio is the wrong input. The work happens once, in the browser, on the file you already have.
Questions
Does auto-tune upload my audio?
No. Apply runs in the tab. The file is not sent to a server.
Is this Auto-Tune Pro or Melodyne?
No. Those are studio products. This page corrects a take to the key and scale you set, then lets you hear the result. It is not a mastering chain, and the name on the button is not a license of the Antares product.
Should I run it on a full mix?
Use a dry vocal. The same pitch shift is applied to the whole file, so drums and chords move with the voice.
What do the presets change?
A preset sets Strength, Retune speed, and Tolerance together. Gentle and Natural leave more of the original motion. Moderate (Default) is 75% / 55% / 25¢. Tight pulls notes in harder. Hard snaps. Strength moves the note center toward the scale. Retune speed is how fast that center arrives, which is also how much vibrato survives. Tolerance is a dead zone: notes already inside it are left alone.
How do I hear the original next to the correction?
After Apply, Load Original and Load Auto-Tuned swap the header player at the same time. Gray on the roadmap is the measured pitch. Pink is the corrected pitch. Click the chart to move the playhead. Save Auto-Tuned as Take writes the correction into a slot A–D and leaves the dry take in its own slot.
Why did a fast run stay wrong?
The pitch track needs a stable period. A grace note or a very short run can be over before that period is found. A slower Retune speed also lets quick ornaments through on purpose. Correction starts when you press Apply. It does not run live while you sing.
The record button in the Controller is grayed out — why?
That means the browser has not been given microphone access for this site yet, or access was denied earlier. Open the permission icon in the address bar, allow the microphone, and reload. You can also skip recording entirely and upload an existing dry vocal file instead.