How to shift voice pitch up or down
- Step 1Drop the voice recording — Add your spoken-word file (MP3, WAV, M4A, OGG, Opus, FLAC). FFmpeg loads in-browser as WebAssembly; the recording is read locally and never uploaded. Pitch shift is a Pro-tier tool, so you need a Pro plan or higher.
- Step 2Decide the direction and amount — Set Semitones: positive raises the voice (younger/lighter/chipmunk), negative lowers it (older/heavier/monster). For a still-human result aim for ±3; for overt characters go further, accepting the formant colouration.
- Step 3Use small steps for subtle disguise — For anonymising a speaker while keeping speech natural, try -2 to -3 or +2 to +3. Use the 0.5 step to fine-tune. Too much and the voice becomes obviously processed; too little and the speaker is still recognisable.
- Step 4Process the clip — Click process. FFmpeg builds
asetrate=<sr*ratio>,aresample=<sr>,atempo=<1/ratio>withratio = 2^(semitones/12)and applies it in one pass, keeping the clip's duration. - Step 5Audition the formant effect — Listen back. Because formants shift with pitch, a +4 voice will sound younger and thinner, a -4 deeper and chestier. If it sounds too cartoonish, pull the semitone value toward 0 and re-run.
- Step 6Download and place in your edit — Output keeps the source format (no separate format picker for this tool). The file is suffixed
-pitched. Because duration is unchanged, drop it straight back into the same slot in your timeline.
Voice character by semitone shift
Approximate perceptual result on spoken voice. Because this is a resampling (non-formant-preserving) shifter, formants move with pitch — that is what drives the character changes below.
| Semitones | Direction | Perceived voice | Stays natural? |
|---|---|---|---|
| +9 to +12 | Up a lot | Chipmunk / cartoon child | No — overt effect |
| +4 to +6 | Up moderate | Younger, lighter, thinner | Borderline |
| +1 to +3 | Up subtle | Slightly higher, still human | Yes |
| 0 | None | Unchanged (re-encoded once) | Yes |
| -1 to -3 | Down subtle | Slightly deeper, still human | Yes |
| -4 to -6 | Down moderate | Older, heavier, chestier | Borderline |
| -9 to -12 | Down a lot | Monster / demon growl | No — overt effect |
The single control and its behaviour
The complete option surface for voice pitch shifting, verified in the tool client.
| Control | Range / step | Effect on voice |
|---|---|---|
| Semitones | -24 to +24, step 0.5 | Retunes the voice; formants move with it |
| Output format | Same as input (no selector) | MP3 in → MP3 out, etc. |
| Duration | Preserved | Lip-sync and edit slots stay aligned |
| Cents | Not available | Use 0.5 semitones (= 50 cents) for fine shifts |
For voice problems this tool does not fix
Pitch shift only retunes. Use these sibling audio tools for the listed problems.
| Problem on the voice | Right tool | Why |
|---|---|---|
| Background noise / hiss | ai-noise-reducer | RNNoise speech neural net, not a pitch op |
| Harsh S sounds | de-esser | Targets sibilance frequencies |
| Uneven loudness | speech-leveler | Evens dialogue level |
| Wrong overall loudness | loudness-normalizer | 2-pass EBU R128 to a LUFS target |
Cookbook
Voice-specific recipes — character creation, disguise, and harmony — with the exact Semitones value and the filter the tool builds. Every clip is processed locally.
Deep narrator / movie-trailer voice
Drop a calm read a few semitones for an authoritative, chestier delivery without re-recording. -3 keeps it human; -5 pushes toward dramatic.
Input: narration.wav (1:20) Semitones: -3 Ratio: 2^(-3/12) = 0.8409 Filter built: asetrate=37083,aresample=44100,atempo=1.1892 Output: narration-pitched.wav (deeper, same 1:20 length)
Chipmunk / kids-show character
A big upward shift produces the classic cartoon voice. The formant shift is the effect, not a flaw.
Input: line.mp3 Semitones: 10 Ratio: 2^(10/12) = 1.7818 Filter built: asetrate=78577,aresample=44100,atempo=0.5612 Output: line-pitched.mp3 (high cartoon voice, same length)
Subtle anonymisation of an interviewee
Two to three semitones down changes recognisability while keeping speech clear and natural. Pair with noise reduction first if the source is noisy.
Input: interview.m4a Step 1 (optional): ai-noise-reducer → clean the speech Step 2: Semitones -2.5 Ratio: 2^(-2.5/12) = 0.8660 Output: interview-pitched.m4a (less recognisable, still intelligible)
Two characters from one actor
Process the same dialogue twice at different shifts to populate a scene with one performer. Identical durations mean both takes drop into the same timeline cleanly.
Source line: "Who goes there?" (0:02) Take A: Semitones +5 → guard-young.wav Take B: Semitones -4 → guard-old.wav Both 0:02 — place back to back in the edit.
Gentle harmony stack on a sung phrase
For a quick vocal-harmony sketch, copy the lead and shift +3 (minor third) and +7 (fifth), then layer. Formant colour is acceptable for a demo; for release-quality harmonies use a dedicated formant-preserving DAW tool.
Lead: vocal.wav (kept) +3 → vocal-3rd.wav +7 → vocal-5th.wav All three same length → stack in DAW for a triad sketch.
Edge cases and what actually happens
Voice sounds like a chipmunk even at a small upward shift
By designFormants are not preserved, so even +4 to +5 thins the voice noticeably. If you want a higher pitch that still reads as the same person, stay within +1 to +3. For larger natural-sounding shifts you would need a formant-preserving algorithm, which this resampling tool is not.
Disguise is too obvious / not enough
ExpectedThere is a narrow window: too few semitones and the speaker is still recognisable, too many and it is clearly processed. Try ±2 to ±3 and the 0.5 step to dial it in. Pitch alone is rarely a robust disguise — combine with noise reduction and EQ for anonymisation if recognisability matters.
Background music in the clip shifts too
ExpectedThe filter shifts the entire mixed signal. If your recording already has music or ambience baked in, that moves with the voice. To shift only the voice, separate the stems before processing — this tool cannot isolate a voice from a mixed track.
Sibilance gets harsher after pitching up
Wrong toolRaising pitch pushes S and T sounds higher and can make them piercing. Run de-esser after the shift to tame them. The pitch shifter itself has no sibilance control.
Clip has noise that becomes weirder after shifting
Wrong toolPitch shifting moves noise along with the voice, sometimes making hiss or hum sound unnatural. Clean the voice first with ai-noise-reducer (RNNoise speech model), then shift.
Set Semitones to 0 expecting a fix
Pass-throughAt 0 the voice is unchanged in pitch but still re-encoded once. This tool only retunes — it does not de-noise, level, or EQ. Pick the matching sibling tool for those tasks.
Free-tier account cannot run it
Requires ProVoice pitch shifting is a Pro-tier feature. Free accounts are blocked. Upgrade to Pro (200 MB / 120 min per file) or higher to process.
Long interview exceeds the duration cap
RejectedPro limits each file to 120 minutes. A multi-hour raw interview is rejected before processing on Pro; Pro-media and Developer remove the duration cap. Trim with audio-trimmer first, or upgrade.
Frequently asked questions
Can I make my voice deeper online for free?
You can lower the pitch in semitones in the browser, but pitch shift is a Pro-tier tool, so a Pro plan or higher is required. -3 to -5 semitones makes a voice deeper while keeping the clip the same length.
Will it still sound like a real person?
Within roughly ±3 semitones, yes. Beyond that, formants move with the pitch (this is a resampling shifter) so the voice becomes a chipmunk going up or a monster going down — great for characters, not for transparent retuning.
Is this a good voice disguise?
It changes recognisability and is useful for casual anonymisation, but pitch alone is not robust against determined identification. Combine it with noise reduction and EQ if anonymity matters.
Does the voice get uploaded?
No. FFmpeg 8.1 runs in your browser as WebAssembly. The recording is processed locally and never sent to a server — appropriate for sensitive or pre-release dialogue.
Can I shift only the voice in a mixed recording?
No. The filter shifts the whole signal, including any music or ambience baked into the file. Separate stems beforehand if you need to shift the voice alone.
Why does pitching up make the S sounds harsh?
Raising pitch moves sibilant frequencies higher, which can sound piercing. Run de-esser after shifting to tame them — the pitch tool has no sibilance control.
How many character voices can I make from one take?
As many as you want — process the same line at different Semitones values (e.g. -4, +5, +9). Each output keeps the original duration, so they all line up in your edit.
What is the smallest shift I can apply?
0.5 semitones (50 cents), the minimum step. That is enough to subtly alter how a voice sits in a mix without an obvious effect.
Does it preserve the timing of my dialogue?
Yes. Duration is preserved by the atempo back-correction stage, so lip-sync and edit slots stay aligned after shifting.
What formats are supported for voice files?
MP3, WAV, M4A/AAC, OGG, Opus, and FLAC. The output stays in the same format as the input — the pitch-shift tool has no output-format selector.
Should I clean the voice before or after shifting?
Clean first. Run ai-noise-reducer (or de-esser) on the raw voice, then pitch shift, so artefacts in the noise do not get pitched into something weirder.
How long a clip can I shift?
Up to your tier's duration limit: 120 minutes on Pro, unlimited on Pro-media and Developer. Free tier cannot run the tool.
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.