Audio Speed & Pitch Changer

Speed and pitch are tied together in nature: speed up a record and the voices squeak. Digital audio can separate them, and there are three useful questions to ask of a recording. Can it be slower without being lower, so a fast passage can be learned? Can it be higher or lower without changing its length, to fit a singer’s key? Or should it behave like a tape, with both moving together? This page answers each one and shows what will happen before you press Apply.

Tape-style: resampling

The simplest method plays the samples at a different rate. Twice the speed halves the length and raises the pitch by an octave, because each wave cycle is squeezed into half the time. The change in semitones is 12 × log2(speed ratio), so 125% raises the pitch by 3.86 semitones and 200% by 12. This is exact and artifact-free, which is why it is the mode of choice for effects such as slowed and reverbed tracks or a deliberate tape stop.

Keeping the pitch: WSOLA

To change the length alone, the sound is cut into overlapping slices of about 46 milliseconds and laid back down closer together (faster) or further apart (slower). Waveform-similarity overlap-add, or WSOLA, chooses where to cut: for each slice it searches a few milliseconds either side of the ideal position for the point where the waveform best continues the previous slice, so the join does not click or cancel. The search runs on a mono mix, and the same positions are used for every channel, which keeps the stereo image intact.

A worked example

Press Sample for a four-second stereo arpeggio. With Speed, keep pitch at 125% (the default), the page shows a new length of 0:03.2 and Apply writes a WAV with 141,120 frames at 44,100 Hz, which is exactly 3.200 seconds (176,400 ÷ 1.25). The notes are still A, C♯, E and A in the same octave. Switch to Pitch only with +12 semitones and the length stays four seconds, within a hundredth of a second, while every note moves up an octave.

Combining speed and pitch

The two controls are independent when the pitch is kept. Internally the sound is first stretched by speed ÷ pitch ratio and then resampled by the pitch ratio, so the length ends up at 1 ÷ speed and the pitch at the semitones you set. For example 150% speed with −2 semitones gives a clip that is two thirds as long and two semitones lower. The line under the sliders shows the resulting pitch change and ratio.

Edge cases

Percussive sound and speech survive time stretching better than sustained notes; a large slowdown of a held chord can sound watery because slices repeat. Very small changes, under about 5%, are close to transparent. Pitch shifting by more than a few semitones changes the character of a voice because its formants move too. Mono and stereo files both work, and the sample rate is unchanged.

Limits

Extreme settings add artifacts, and there is no formant correction, transient detection or beat matching. The tool does not know the tempo or key of the music. Output can be WAV, Opus or AAC, but not MP3, and the processing happens on your device.

Related tools

• Sample Pitch & Time-Stretch Calculator works out the numbers for a pitch or tempo match.
• Audio Cutter isolates the passage you want to practise.
• Reverse Audio plays a clip backwards.
• Metronome keeps time while you practise at the new speed.