How to Fix Audio Distortion

Each distortion has a signature. (1) Digital clipping: harsh, buzzy crackle riding only on loud transients (drum hits, consonants); waveform shows flat tops at full scale. (2) Analogue / preamp saturation: warm, slightly fuzzy 'tube' character on loud notes; gradual rather than abrupt. (3) Microphone overload: similar to clipping but baked in before the converter — a 'splatty' or 'compressed' quality on high-SPL sources (drums close-miked, brass, shouting). (4) Codec artefact: swirly, watery, metallic quality especially on cymbals, sibilants, sustained vowels — present even at quiet sections. (5) Ground hum / RFI: tonal buzz at 50/60 Hz with harmonics, present in silence and in signal alike. Listen with headphones at moderate volume; identify the loudest distortion type first, because masking can hide secondary problems until the primary is fixed.

Cause: signal exceeded 0 dBFS at the converter, in a software fader, or in a lossy codec encode. Identification: visible flat tops at full scale in Audacity (View > Waveform; zoom in until samples are visible). Fix: iZotope RX 11 De-clip handles brief, isolated clips well — set Algorithm A for sustained content, Algorithm B for transient hits. Audacity Effect > Clip Fix is the free path. Adobe Audition Diagnostics > DeClipper. Repair quality scales inversely with clip duration: single-sample clips are fully recoverable; 100 ms of sustained clipping is unrecoverable. Prevent by gain staging: peaks at -6 to -12 dBFS during tracking; 24-bit recording for headroom; true-peak limiter on the master with -1 dBTP ceiling. See [fix-audio-clipping](/guide/fix-audio-clipping) for the dedicated workflow.

Causes: source SPL exceeds the mic's maximum SPL rating (e.g., a 132 dB SPL screaming vocal hitting an SM7B's 180 dB rating is fine; but a budget condenser rated 120 dB SPL clips on the same source). Or the preamp gain is set so high that the input stage saturates before the converter ever sees the signal. Identification: distortion present even when the converter shows safe levels (-12 dBFS); waveform shows asymmetric squashing rather than symmetric flat tops. Fix: re-record with the pad switch engaged (most condensers have a -10 or -20 dB pad), preamp gain reduced, or the mic moved further from the source. In post: nothing fully repairs preamp distortion — it is harmonic content added before the digital domain. iZotope RX De-distort can reduce some low-grade preamp saturation, but the cure is at-source. Document gain settings with each take to identify the failure mode for next session.

Causes: lossy encoding at insufficient bitrate. MP3 below 96 kbps, AAC below 64 kbps, Opus below 24 kbps all produce audible artefacts: pre-echo (a brief 'whoosh' before transients), high-frequency smearing (cymbals lose detail), and metallic ringing on sustained vowels and brass. Identification: the artefact is consistent across the entire file at the same frequencies and timbres; spectrogram (Audacity: View > Spectrogram) shows bands of energy missing above some cut-off (often 16 kHz for low-bitrate MP3). Fix: re-encode from a higher-quality source. Going from 64 kbps MP3 to 320 kbps MP3 does not improve quality — the discarded data is gone. Convert via [MP3 to WAV](/convert/mp3-to-wav) and re-encode at higher bitrate only if the original source is also lossless. Prevent: encode lossy formats once, at the final delivery stage, from a lossless master. Never transcode lossy to lossy if avoidable.

Causes: ground loops, USB power noise, cellular RF picked up by unbalanced cables, fluorescent lighting, dimmer switches. Identification: tonal buzz at 50 or 60 Hz with harmonics at 100/150/200 (or 120/180/240); present during silence; doesn't change with signal level. Fix: prevention is all single-circuit power, balanced cables, distance from EMI sources — see the dedicated [fix-audio-hum](/guide/fix-audio-hum) guide. Post-production: notch filters at the fundamental and harmonics, or iZotope RX De-hum (one-click). For low-grade hum below the noise floor, ignore it; for hum that is loud enough to fail QC, the only options are notch filtering (which dips the bottom octave) or re-recording.

Sample-rate conversion done with a low-quality resampler introduces aliasing — high-frequency content folds back into the audible band as inharmonic distortion. Symptoms: faint 'sparkle' or 'shimmer' artefacts above 8 kHz on resampled material. Bit-depth reduction without dither produces quantisation noise on quiet passages — a granular hiss that increases with attenuation. Both are avoidable: use ffmpeg's high-quality resampler ('-af aresample=resampler=soxr -ar 44100'), or sox itself, both excellent. For bit-depth reduction, apply triangular dither: 'ffmpeg -i input.wav -c:a pcm_s16le -af aresample=osf=s16:dither_method=triangular output.wav'. AudioUtils' converters use ffmpeg with high-quality SOX-based resampling — sample-rate conversions through the browser tools do not introduce audible aliasing.

Some distortion is permanent; the right move is to accept it and redirect time. Heavy clipping over more than 5% of file duration: residual artefacts will be audible regardless of plugin. Codec artefacts in a file you don't have a higher-quality source for: nothing recovers the discarded audio. Long sustained mic overload on a long-form recording: unrecoverable. Background noise that exceeds the signal level in some passages (busy outdoor location, recorded with a single mic at distance): no plugin will pull intelligible voice out of noise that loud. The rule: spend a maximum of 15 minutes on restoration of any one defect; if it is not acceptable by then, decide between live-with-it, re-record, or replace from a different take. The opportunity cost of restoration is sometimes greater than the cost of fresh capture.