Signal Processing
What Is Audio Artifacting?
Quick answer
What artifacts are
An artifact in audio is any sound present in the output that wasn't in the input. When compression algorithms make decisions about what to discard — or when processing tools make incorrect assumptions about the audio — they sometimes introduce sounds that weren't there originally. These are artifacts.
They're distinct from noise, which is random and broadband. Artifacts have specific character related to the processing that created them. You can often identify the cause of an artifact by what it sounds like.
Common codec artifacts
| Artifact | Sounds like | Cause | When it appears |
|---|---|---|---|
| Pre-echo / pre-ringing | A faint ghost of a sound audible just before it occurs | Transform boundary effect in MDCT; energy "leaks" from a loud event into preceding frames | Sharp transients at low bitrates (cymbals, piano notes) |
| High-frequency smearing | Cymbals sound blurry; sibilance becomes a soft hiss | Aggressive high-frequency cutoff in the encoder at lower bitrates | Below 192 kbps MP3; complex material |
| "Underwater" / "metallic" sound | Audio sounds like it's behind glass or being played through a filter | Extreme bitrate reduction; the psychoacoustic model makes poor decisions on complex audio | Below 96 kbps; dense music |
| Quantization noise | Low-level broadband hiss that rises and falls with the signal | Limited bit depth; each sample rounded to the nearest available value | Low bit depth (8-bit); rarely audible at 16-bit |
| Pumping / breathing | Volume pulses rhythmically or breathes in and out | Over-compressed codec or heavy noise gate | Low bitrate encoding of music; heavy noise reduction |
Pre-echo: the most counterintuitive artifact
Pre-echo is worth explaining in more detail because most people have never heard it named — even though they've almost certainly heard it in low-quality MP3s.
Lossy codecs process audio in blocks of time (frames). When a sudden loud sound — a sharp piano note, a cymbal crash — occurs in a frame, the encoder has to allocate bits for it. At low bitrates, the energy from that loud sound can "bleed" into the quiet portion just before it within the same frame. The result: you hear a faint, reverb-like shadow of the sound arriving slightly before the sound itself.
Pre-echo is particularly noticeable on solo piano and acoustic guitar recordings at low bitrates — two instrument types that have sharp transients (the attack of the note) followed by clear decay. Listen for a subtle "woosh" just before each note strike in low-quality audio.
Noise reduction artifacts
Noise reduction tools have their own distinctive artifact signature. When the reduction is applied too aggressively, the algorithm starts treating parts of the desired signal as noise and attenuating them. The result is a warbling, metallic shimmer — the audio seems to flutter or oscillate, like a corrupted radio signal or a reverb trail that has been electronically filtered.
The artifact is created in the frequency domain: the spectral subtraction algorithm creates modulation in how frequencies vary over time. Mild cases sound like a light reverberation that shouldn't be there. Severe cases make speech sound like it's being processed through a vocoder.
How to avoid codec artifacts
- Use an appropriate bitrate:192 kbps MP3 eliminates most audible codec artifacts for music. 128 kbps is acceptable for voice. Below 96 kbps, artifacts are consistently audible on complex material.
- Choose the right codec:AAC produces fewer artifacts than MP3 at the same bitrate. Opus is dramatically better at low bitrates. For quality-critical work at lower bitrates, use a modern codec.
- Encode from a lossless source:Encoding from a lossy source (re-encoding MP3 to AAC) compounds artifacts. Always encode from the original lossless source when possible.
- Limit noise reduction intensity:Apply the minimum noise reduction needed, not the maximum available. Use the preview function to check for artifacts before committing.
Last updated: March 28, 2026