Change audio speed

Keeping voices natural

Playing audio faster the naive way raises its pitch, because speed and pitch are the same thing when you simply play a waveform quicker. That is the chipmunk effect, and it makes sped-up speech unpleasant rather than efficient.

Keeping the pitch natural uses time-stretching instead — the audio is broken into short segments and re-timed, leaving the frequencies where they were. This is what podcast apps do at 1.5x, and it is on by default here.

Turning it off deliberately

The pitch-shifting behaviour has real uses: dropping a voice for anonymity, matching a sample to a key, or the comic effect. Turn off keeping the pitch and you get the tape-speed result.

Extreme speeds work

The time-stretching filter handles half to double speed in one pass. Anything beyond that is chained automatically, so 3x and 4x work properly rather than being capped or quietly breaking — which is where a lot of tools stop.

Where the quality limit is

Time-stretching is very good between about 0.5x and 2x. Push much beyond that and speech develops a faint metallic quality, because the algorithm is repeating or discarding more of the waveform than it can hide. It stays intelligible, which is usually the point at 3x.

Duration

Changes proportionally — a 30 minute recording at 1.5x is 20 minutes. Both figures are reported, which helps when you are fitting a time slot.

Where the time actually goes

Worth a sanity check before reaching for this: for a long recording, trimming the dead air at the start and end often saves more time than speeding up the whole thing, and costs no quality at all. A lecture that begins with four minutes of people arriving is better trimmed than accelerated.

Related

Change video speed does the same thing with the picture in step.