Two different things people mean by “faster”
Changing playback speed sounds like one operation but is really two, and picking the wrong one gives you a result you did not want.
Keeping the pitch means the track finishes sooner while every voice and instrument stays at the frequency it was. This is what podcast apps do at 1.5×, and it is the only sensible option for speech — a lecture sped up the other way is unlistenable within about ten seconds.
Changing the pitch replays the entire waveform at a different rate. Everything rises together, exactly as it does when a record player runs fast or a tape is wound on. This is the chipmunk effect, the nightcore effect, and the honest way to emulate analogue equipment.
How pitch-preserving time-stretching works
Holding pitch constant while changing duration is a genuinely hard problem, because in a raw waveform the two are the same thing. The standard approach — and the one FFmpeg's atempo filter uses — is to cut the audio into short overlapping windows and then space those windows further apart or closer together, crossfading between them.
The frequencies inside each window are untouched, so pitch survives, but the seams are where the artefacts live. At moderate settings they are inaudible. Pushed hard, sustained tones can pick up a faint warbling or a doubling that sounds like a very short echo. Speech tolerates this unusually well because it is full of natural discontinuities; a solo instrument is where you hear it first.
One implementation detail worth knowing: a single atempo pass only accepts factors between 0.5 and 2. Anything outside that is applied as a chain of passes, so 4× is two 2× stages. That is why the extremes accumulate slightly more character than the middle of the range.
Choosing a speed
- 1.25× – 1.5× — the sweet spot for lectures and podcasts. Almost no artefacts and a real time saving.
- 1.75× – 2× — workable for familiar speakers and clear recordings. Harder with accents or poor audio.
- 0.75× — transcription, language learning, and working out a fast musical passage.
- 0.5× and below — detailed analysis, where artefacts matter less than hearing every detail.
Common uses
- Getting through a recorded lecture or meeting in less time
- Slowing a song down to learn a part by ear
- Making a track fit an exact runtime for video
- Producing nightcore or vaporwave edits
- Slowing speech for transcription or language practice