How to Fix Audio Clipping

Three independent checks confirm clipping. Visual: zoom into the waveform view in Audacity, Reaper, or your DAW until samples are visible — clipped sections have flat tops at exactly 1.0 (full scale) instead of rounded peaks. Metered: peak meters that hit and stay at 0 dBFS during transients indicate clipping; meters with a 'CLIP' indicator latch on overload. Aural: clipping sounds like a crackle or fuzz layered on top of loud syllables, drum hits, and brass notes — the harshness rides only with loud peaks, never on quiet sections. Audacity has Analyze > Find Clipping which annotates every clipped run; Adobe Audition's Statistics panel reports possibly-clipped sample counts. Convert any lossy source to [WAV](/convert/mp3-to-wav) before deep analysis — MP3 frame compression hides single-sample clips.

The two sound and look different. Analogue clipping happens in the gain stages of a hardware preamp, microphone, or tape circuit. Soft clipping rounds peaks gradually, adding harmonic saturation many engineers find musical — guitar amps and tape machines exploit it deliberately. The waveform shows a slightly squashed peak rather than a hard flat top. Digital clipping happens at the converter or in a software fader exceeding 0 dBFS. The truncation is mathematically abrupt: every sample over 1.0 becomes 1.0 exactly. The resulting waveform is flat-topped, the spectral signature is harsh odd-harmonic distortion, and the artefact is permanent in the file. Lossy codecs can also create 'inter-sample peaks' that clip during D/A conversion even when the digital samples sit just below 0 dBFS — see the headroom rule below.

The single most important habit: leave headroom. Set the input gain so that the loudest expected source peaks at -6 to -12 dBFS on the meter, never -3 dBFS or hotter. 24-bit recording gives 144 dB of dynamic range — even peaking at -18 dBFS leaves you 126 dB above the noise floor, plenty for any source. For unpredictable sources (live music, interviews, sound effects capture), aim even lower, around -18 dBFS, and let normalization in post bring the level up. Use the pad switch (-10 or -20 dB) on the audio interface for loud sources like drums or trumpet at close range. Some interfaces (RME, Apogee, MOTU) include a clip indicator that latches even on a single overload — leave it visible during the take and adjust gain whenever it lights.

Inside a DAW, internal busses run at 32-bit float with effectively unlimited dynamic range — clipping a fader inside the mix does not damage the audio because the bus can represent values above 0 dBFS losslessly. But the master bus output, the bounce file, and any 16/24-bit integer export will clip on every sample that exceeds 0 dBFS. Best practice: keep the mix bus peaking at -6 to -3 dBFS; add a brick-wall [limiter](/guide/what-is-audio-limiter) with ceiling -1 dBTP before the master meter; export to 24-bit WAV first, then derive lossy distribution copies via [WAV to MP3](/convert/wav-to-mp3). The -1 dBTP ceiling guards against inter-sample peaks introduced by lossy codec decoding on the listener's side.

Software declippers reconstruct the missing peak by interpolating across the flat region from samples immediately before and after. Quality scales with the duration of the clip: a single-sample truncation is fully recoverable; a 100-sample flat-top is a guess; a 5-second sustained overload is unrecoverable. iZotope RX 11 De-clip is the industry standard — set Threshold to detect the flat tops, choose 'Algorithm B' for fast transients (drums, plucked strings) or 'Algorithm A' for sustained content (vocals, brass). Adobe Audition: Effects > Diagnostics > DeClipper. Audacity (free): Effect > Clip Fix; works on isolated transient clipping but is weaker than RX. Apply declipping before any other processing so subsequent EQ and compression operate on the repaired signal. Save the repair pass as a new file — never overwrite the original capture.

Repair has limits. As a rule of thumb: clipping that appears on more than 5% of total file duration cannot be fully restored — declipping artefacts will be audible regardless of plugin quality. Sustained clipping on long notes (held vocals, organ chords, sustained guitar) is unrecoverable; the waveform's natural shape during the held section is unknowable. Voice-over for ACX or broadcast must be free of audible clipping artefacts; declipped takes will be flagged. If the source can be re-captured (you still have the talent in front of the mic, or the music project is still tracking), 30 seconds of re-recording with corrected gain beats two hours of restoration work. Document the gain settings that caused the clip so the same source does not clip on the retake.

Converting to lossy formats can introduce clipping that was not in the source. The MP3 and AAC encoding processes can produce inter-sample peaks 0.5-1 dB higher than the input file's digital peaks. A 0 dBFS WAV may decode to a clipping MP3. Solution: master to -1 dBTP true-peak ceiling using a true-peak limiter (FabFilter Pro-L 2, iZotope Ozone Maximizer, or free TDR Limiter 6 GE). When converting via [WAV to MP3](/convert/wav-to-mp3) or [WAV to AAC](/convert/wav-to-aac), always start from a -1 dBTP master so the encoder has 1 dB of safety. Loud over-mastered sources (mastered to 0 dBFS for streaming volume wars) are the most common cause of clipping in delivered MP3s — the encoder reproduces the input's lack of headroom faithfully and adds inter-sample peaks on top. See [fix audio distortion](/guide/fix-audio-distortion) for the related distortion pathway.