Audio Pitch Shifter

Move audio to a different pitch and choose what that does to the character of the sound. Everything below is measured on the reference machine rather than claimed.

• Two octaves either way, to the cent. Twenty-four semitones up or down, with a hundred cents of fine tuning on top. The pitch you get lands within 0.08 cents of the pitch you asked for across the whole range.
• Voices keep their character. The resonances that make a voice recognisable are held still while the notes move, measured to within 3% of where they started at a full octave shift.
• The length does not drift. A pitch shift leaves the duration exactly as it was. Only Tape mode changes it, and only because that is what Tape mode is for.
• Timbre as its own control. Move vocal character up or down an octave without moving the pitch at all — for disguising a voice, or matching two speakers recorded differently.
• It finds the key for you. Detect the key, pick the key you want, and the shift is worked out for you instead of counting semitones by hand.
• Nothing is uploaded. The audio is decoded, shifted and re-encoded on your device. No file data crosses the network, so there is nothing to wait in line for.

A voice is two things at once. The notes come from the vocal folds, and the character comes from the shape of the throat and mouth, which sculpts that raw buzz into peaks of emphasis at fixed frequencies. Those peaks are what tell you whose voice you are hearing, and — this is the important part — they do not move when someone sings a higher note. A soprano and a bass singing the same word have their peaks in roughly the same places.

• Natural voices moves the notes and leaves the character peaks where they are. A vocal shifted up sounds like the same person singing higher. This is the setting for singing, speech, podcasts and anything with a person in it.
• Follow the pitch moves everything together, which is what speeding up a tape does. The result is the chipmunk effect going up and the monster effect going down. Correct for instruments — a piano really does resonate differently note to note — and correct when the effect is the point.
• Tape moves pitch, character and length together, exactly like a physical machine running fast or slow. The only mode that changes how long the file is.

Most online pitch tools only offer the middle one, which is why their output has a reputation for sounding artificial. The difference is immediately audible on any vocal — press Audition and switch between them before committing to an export.

The real task is almost never "shift this by four semitones". It is "this backing track is in F and my singer needs D", and working out the number in between is not something you should be doing by hand. Press Detect key and the tool reads the track, folds every note it hears onto the twelve pitch classes and compares that against the fingerprint of all twenty-four major and minor keys.

• It shows its confidence. A number, not a bare answer, so you can tell a certain reading from a coin toss.
• It names the runner-up when it matters. A major key and its relative minor contain exactly the same seven notes, so that pair is genuinely ambiguous. When the two are close the tool says so instead of picking one and sounding sure.
• It takes the short way round. C to A is three semitones down, not nine up. Both land on A, but only one of them leaves the singer where they can still reach the notes.
• You can always override it. The detected key fills in the slider; it never applies a shift on its own. Key detection is a good heuristic, not an oracle, and a tool that silently transposed by a wrong guess would be worse than one that never guessed.

• Small shifts sound better than large ones. Anything within about five semitones is usually indistinguishable from a real performance at that pitch. Past an octave you are asking the tool to invent detail that was never recorded, and it will start to sound processed however good the maths is.
• Use cents to fix tuning, semitones to change key. A recording made a little sharp needs twenty or thirty cents, not a whole semitone. The two sliders are one axis: a semitone is a hundred cents.
• Pick Natural voices for anything with a person in it. Including rap, spoken word and backing vocals. Pick Follow for solo instruments, drums and sound effects.
• Export lossless if you are going to keep editing. FLAC and WAV come back with no generation loss. MP3 at 192 kbps is fine for listening, but shifting an MP3 and then shifting the result again stacks two lots of compression.
• Leaving the pitch at zero is not a wasted run. If you have not moved the pitch and you ask for the same format back, the file is copied rather than re-encoded. It is instant and bit-for-bit identical, which is a genuinely useful way to check a file round-trips.

• Transposing a backing track. Move a karaoke or practice track into a singer's comfortable range without it sounding processed.
• Matching a sample to a project. Nudge a loop or one-shot into the key of whatever you are building.
• Disguising a voice. Shift the pitch down and the timbre separately, which sounds far less obviously altered than pitch alone.
• Fixing a tuning drift. Correct a recording made on equipment running slightly fast or slow, in cents rather than semitones.
• Practising an instrument. Bring a recording down to a key you can play along with while you learn it.
• Making a character voice. Follow mode up for a chipmunk, down for a monster, and Tape mode for an old-machine effect.

Powered by optimised audio processing algorithms.

• Audio speed changer
• Audio equalizer
• Audio compressor
• Audio cutter
• Volume booster
• Audio converter