Pitch Shifter
Move the pitch, leave the length exactly where it was.
or paste one from your clipboard
How to use it
- Drop an audio file onto the page. It is decoded in the tab; nothing is sent anywhere.
- Drag the semitone slider. Each step is one half step on a piano. 12 steps is a full octave, and the interval name under the slider tells you what you have asked for.
- If you need to land between the notes, nudge the cents control. One hundred cents makes a semitone, so ±10 cents is about the smallest change most people can hear on a sustained note.
- Click Preview the shift to render it and hear it in the player. Pitch shifting is heavy enough that it is not recomputed on every drag.
- Choose a format and click Download. The length of the file is identical to what you started with.
Why pitch and length are normally locked together
A digital recording is a list of numbers plus a rate at which to read them. Read the list faster and everything rises; read it slower and everything falls. That is the only pitch control tape and vinyl ever had, and it is why a sped-up recording always sounds like a smaller person in a smaller room. The two properties are welded together by the format itself.
To separate them you have to rebuild the signal. This tool stretches the audio in time first. Cutting it into overlapping windows of about 46 milliseconds and laying them back down at wider spacing, searching for the offset where each window best continues the waveform already written. That gives a longer file at the original pitch. Then it resamples the stretched version by the same ratio, which drags the length back to where it started and takes the pitch up with it. The length changes cancel; the pitch change does not. Shift down and the same trick runs in reverse.
Reading the numbers
Pitch is logarithmic, so intervals are ratios rather than differences. One semitone is a ratio of the twelfth root of two, about 1.0595. Stack twelve of them and you get exactly 2, which is why +12 doubles every frequency and lands you an octave up. It also means the same slider position sounds like the same musical distance whether you are working on a bass line or a cymbal.
Cents divide each semitone into a hundred. They exist because tuning problems are usually far smaller than a half step: a string that has drifted, a sample that was recorded at 44,100 Hz and played back at 48,000, a singer who sat consistently a touch under the note. Below roughly 5 cents almost nobody hears a difference. Between 10 and 25 cents most people hear "slightly off" without being able to say which way. Past 50 cents you are halfway to the next note and the ear starts rounding to it.
Where it breaks down
Two things limit how far you can push a shift. The first is stretching artefacts: at large ratios the algorithm has to reuse or discard a lot of material, and sustained tones develop a faint warble while transients lose their edge. The second, and the one people notice without knowing why, is formants. A voice has fixed resonances set by the size of the throat and mouth, and those resonances stay put when a real singer changes note. Shift the whole spectrum and the resonances move too, so the listener hears a body that changed size mid-sentence.
Keep shifts small when you want the result to pass as unprocessed. Go large deliberately when you want an effect. A voice dropped an octave for a character, or raised for a chipmunk part. And let it sound like the effect it is.