You speed up a video to 2x and suddenly everyone sounds like a chipmunk. The deeper someone's natural voice, the more ridiculous the pitch shift sounds. This is not a software glitch. It is a fundamental property of how sound waves work.
Why speed and pitch are linked
Sound is a vibration. The pitch you hear is determined by how many times per second the air vibrates at your eardrum, measured in Hertz (Hz). A typical adult male voice has a fundamental frequency around 85 to 180 Hz. A typical adult female voice is around 165 to 255 Hz.
When you play audio at double speed, you are compressing the waveform in time. Every cycle of vibration that originally took one second now happens in half a second. That means the frequency doubles. A 150 Hz voice becomes 300 Hz. In musical terms, doubling the frequency raises the pitch by exactly one octave. A man's voice at 2x speed sounds roughly like a woman's voice. At 3x, it sounds cartoonish.
This is the same principle behind vinyl records: play a 33 RPM record at 45 RPM and the music sounds faster and higher-pitched.
The three approaches to fixing it
Approach 1: Mute the audio entirely. This is the simplest and most common solution, especially for timelapse or montage sequences. Remove the original audio and replace it with background music that was not affected by the speed change. This works for any speed multiplier and produces clean results.
Approach 2: Time-stretching with pitch correction. Professional audio processing can speed up audio while preserving the original pitch using algorithms like WSOLA (Waveform Similarity Overlap-Add) or phase vocoder techniques. These algorithms chop the audio into tiny overlapping segments, discard segments to shorten the duration, and crossfade the remaining segments together so the pitch stays the same.
FFmpeg's atempo filter does this automatically. When you speed up a video using MediaBrew's speed tool, the atempo filter adjusts the audio speed while attempting to minimize pitch distortion. At 2x, the results are reasonably natural. At 3x and above, artifacts like metallic ringing or syllable smearing start to appear because the algorithm is discarding too much audio data.
Approach 3: Record separate audio. If you are creating a tutorial or narration where audio quality matters and you need a sped-up visual, record the voiceover separately at normal speed. Then combine the normal-speed voiceover with the sped-up video using the add audio tool. This gives you full control over both elements.
Why the chipmunk effect sounds worse at higher speeds
At 2x, pitch shifts up by one octave. Recognizable but tolerable. At 3x, pitch shifts up by about 1.5 octaves. Voices become squeaky. At 4x, pitch shifts up by two octaves. A 150 Hz male voice becomes 600 Hz, which is higher than most female singing voices. Speech becomes nearly incomprehensible.
The human ear is finely tuned to the frequency range of speech (roughly 85 Hz to 3000 Hz for the fundamental and primary harmonics). When pitch-shifted voices move outside the expected range, our brains flag the audio as unnatural, which is why it sounds funny rather than just "faster."
How FFmpeg's atempo filter works
The atempo filter in FFmpeg adjusts audio speed while preserving pitch. It has a valid range of 0.5x to 100x. For speed multipliers outside the 0.5-2.0 range, FFmpeg chains multiple atempo filters together. For example, 4x speed is achieved by applying atempo=2.0 twice (2.0 * 2.0 = 4.0).
The filter works by analyzing the audio in small windows, finding similar overlapping regions, and splicing them together at the new speed. At moderate speed changes (up to 2x), this produces natural-sounding results. At extreme speeds (8x, 16x), the audio quality degrades significantly, which is another reason to mute audio for fast timelapses.
The practical recommendation
For speeds up to 1.5x: Keep the original audio. The pitch shift is minimal and the atempo filter handles it well. For 2x: The audio is usable but noticeably altered. Consider keeping it for casual content or muting for polished content. For 3x and above: Mute the original audio and add music. The pitch distortion and processing artifacts make the original audio more distracting than helpful.
The MediaBrew speed tool applies the atempo filter automatically when you change speed, so the audio pitch is corrected by default. If you prefer to strip the audio entirely, use the mute tool first or after the speed change.