How to pitch correct a recording in your browser
- Step 1Identify how far off the recording is — Work out roughly how flat or sharp the take is in cents or semitones — by ear against a reference, or with a tuner. A typical 'sounds a bit flat' is around 20–50 cents (0.2–0.5 semitone). Remember the shift is global, so this only works if the whole take is off by the same amount.
- Step 2Load the recording — Drop the file (WAV/FLAC preferred for corrections, but MP3/M4A/OGG/Opus work). FFmpeg loads in-browser as WebAssembly; the take is read locally and not uploaded. Pitch correction here is a Pro-tier tool.
- Step 3Enter a small Semitones value — Use Semitones with a small value: +0.5 to raise a flat take, -0.5 to lower a sharp one. The step is 0.5 (50 cents). For a take that is about a semitone off, use ±1. Stay within ±3 to keep it transparent.
- Step 4Apply the correction — Click process. FFmpeg builds
asetrate=<sr*ratio>,aresample=<sr>,atempo=<1/ratio>and applies it in one pass, preserving the take's duration. - Step 5Check against the reference pitch — Play the corrected take against your reference (tuner, other instruments, click in key). If still off, adjust by another 0.5 and re-run from the original — re-correcting an already-corrected file stacks re-encodes.
- Step 6Download the corrected take — Output keeps the source format and is suffixed
-pitched. Because the length is unchanged, drop it back into the same position in your session.
Correction amount reference
Map a 'how far off' estimate to a Semitones value. The step is 0.5 semitone (50 cents), so this is the practical correction grid.
| Recording is | Direction to fix | Semitones | Naturalness |
|---|---|---|---|
| ~50 cents flat | Raise | +0.5 | Transparent |
| ~50 cents sharp | Lower | -0.5 | Transparent |
| ~1 semitone flat | Raise | +1 | Transparent |
| ~1 semitone sharp | Lower | -1 | Transparent |
| 2–3 semitones off | Raise/lower | ±2 to ±3 | Mostly natural |
| More than 3 semitones off | Raise/lower | ±4+ | Audible colouration |
What this correction can and cannot do
Scope of a global pitch shift versus what you might expect from autotune.
| Task | This tool? | Where to go instead |
|---|---|---|
| Whole take uniformly flat/sharp | Yes — small semitone shift | — |
| Per-note pitch correction (autotune) | No — global only | DAW pitch-correction plugin |
| Pitch drifts within the take | No — single fixed shift | Re-record or DAW automation |
| Recording too quiet/loud | No | loudness-normalizer |
| Background noise on the take | No | ai-noise-reducer |
Correction control and limits
Verified control surface and tier limits.
| Item | Value |
|---|---|
| Control | Semitones, -24 to +24, step 0.5 |
| Finest step | 0.5 semitone = 50 cents |
| Output format | Same as input |
| Duration | Preserved |
| Min tier | Pro (200 MB / 120 min / 10 files) |
Cookbook
Small-correction recipes for flat/sharp takes, with the exact Semitones value and the filter the tool builds. All processing is local; nothing uploads.
Raise a take that is 50 cents flat
A vocal or guitar take that sits half a semitone low. +0.5 brings it up to pitch with no audible colouration, keeping the take's length.
Input: take.wav (44100 Hz, ~50 cents flat) Semitones: 0.5 Ratio: 2^(0.5/12) = 1.0293 Filter built: asetrate=45392,aresample=44100,atempo=0.9715 Output: take-pitched.wav (≈ in tune, same length)
Lower a sample recorded slightly sharp
A field-recorded instrument is about a semitone sharp against the project. -1 brings it down; small enough to stay transparent.
Input: field-sample.flac (~1 semitone sharp) Semitones: -1 Ratio: 2^(-1/12) = 0.9439 Filter built: asetrate=41617,aresample=44100,atempo=1.0595 Output: field-sample-pitched.flac (down a semitone, same length)
Match a guitar tuned a quarter-tone low to a fixed-pitch instrument
The guitar drifted ~25 cents flat. The finest available step is 0.5 semitone (50 cents), so a perfect 25-cent fix is not possible — choose 0 (leave) or +0.5 (slightly over). For sub-quarter-tone precision, retune and re-record.
Input: guitar.wav (~25 cents flat)
Options: Semitones 0.5 over-corrects (+50 cents)
Semitones 0 leaves it (-25 cents)
Neither lands exactly — finest step is 0.5 semitone.Bring two slightly mismatched takes into agreement
Comp take B is a touch flatter than take A. Estimate the gap (~0.5 semitone) and raise take B by 0.5 so both sit together. Both stay the same length for comping.
Take A: reference, in tune Take B: ~50 cents flatter → Semitones +0.5 Ratio: 1.0293 Output: takeB-pitched.wav — now matches take A
Why a per-note problem needs a different fix
If the singer is flat on the chorus but sharp on the verse, a single global shift cannot fix both. This example shows the limitation rather than a fix.
Verse: +30 cents (sharp) Chorus: -40 cents (flat) Global shift of +0.5 helps the chorus but worsens the verse. → Use a DAW per-note pitch-correction plugin instead.
Edge cases and what actually happens
Singer is flat on some notes, sharp on others
Not supportedThis is a uniform global shift, not autotune. It moves every note by the same amount, so it cannot correct a take that is flat in places and sharp in others. Use a per-note pitch-correction plugin in a DAW for that.
Need to correct by less than a quarter-tone
ExpectedThe minimum step is 0.5 semitone (50 cents). A 20–30 cent correction cannot be applied exactly — you would over- or under-shoot. For sub-quarter-tone precision, retune the source and re-record, or use a DAW tuner.
Pitch drifts gradually across the take
Not supportedA single Semitones value applies one fixed shift to the whole file. If the take starts in tune and drifts flat, no single value fixes it. This needs time-varying correction (DAW automation) or a re-record.
Correction makes the take sound processed
ExpectedWithin ±3 semitones the resampling colouration is minimal. If a correction sounds artificial, you are probably shifting too far — meaning the take was off by more than a small amount, which is usually better re-recorded than corrected.
Re-correcting an already-corrected file
Quality lossEach pass re-encodes (and re-resamples). Correcting a file twice stacks artefacts. Always go back to the original take and apply a single combined Semitones value rather than chaining corrections.
Take has noise or hum that you also want gone
Wrong toolPitch correction only retunes. For noise on the take, run ai-noise-reducer (RNNoise) first; for level issues use loudness-normalizer or speech-leveler.
Lossy take corrected and re-encoded
Quality lossAn MP3 take comes out as MP3, re-encoded once. For correction work, export the original take to WAV first, correct, then convert to MP3 at the end if needed — avoids a second lossy generation on top of the correction.
Free-tier user wants to correct a take
Requires ProPitch correction via this tool is a Pro-tier feature; free accounts cannot run it. Pro allows 200 MB / 120 min per file. Upgrade to correct takes.
Frequently asked questions
Can I pitch-correct a recording in the browser?
Yes, for uniform corrections: enter a small Semitones value to shift the whole take up or down. It runs in-browser via FFmpeg WebAssembly. This is a Pro-tier feature, and it is a global shift, not per-note autotune.
My take is slightly flat — what value do I use?
Roughly 50 cents flat → +0.5 semitones; about a semitone flat → +1. The step is 0.5 (50 cents), which suits most small corrections. Sharp takes use negative values.
Is this autotune?
No. It applies one fixed shift to the entire recording. It cannot correct individual notes that are off by different amounts — for that you need a per-note pitch-correction plugin in a DAW.
Will correcting the pitch change the length of my take?
No. The atempo stage preserves duration, so the corrected take is exactly as long as the original and slots back into the same place in your session.
How precise can the correction be?
Down to 0.5 semitone (50 cents). There is no cents field, so corrections smaller than a quarter-tone are not possible — you would over- or under-shoot.
Will a small correction sound natural?
Yes. Within ±3 semitones the resampling colouration is minimal, so a corrected take still sounds like the original performance.
Can it fix pitch that drifts during the take?
No. A single Semitones value shifts everything by the same amount. A drifting take needs time-varying correction (DAW automation) or a re-record.
Is my recording uploaded?
No. FFmpeg 8.1 runs in your browser as WebAssembly, so raw takes and session files are processed locally and never sent to a server.
Should I correct before or after other processing?
Clean (noise, level) on the original first if needed, then apply a single pitch correction. Avoid chaining corrections — go back to the original and combine into one Semitones value.
What format does the corrected file use?
The same as the input. There is no output-format selector for this tool, so the container is preserved. Use WAV/FLAC for correction work to avoid extra lossy generations.
What if the take is way off pitch?
Shifts beyond ±3 semitones start to colour the sound audibly. If a take is that far off, re-recording usually beats correcting. This tool is best for small, global nudges.
What tier do I need?
Pro or higher. Free tier cannot use the tool. Pro handles 200 MB / 120 min per file; Pro-media and Developer remove the duration cap.
Privacy first
Every JAD Audio tool runs entirely in your browser via FFmpeg (WebAssembly) and RNNoise. Your audio files never leave your device — verified by zero outbound network requests during processing.