Clipping detector

How clipping is detected

Digital audio cannot go above full scale (0 dBFS, a sample value of 1.0). When a recording, plug-in or export pushes the signal past it, the tops of the waveform are cut flat. That sounds like crackle or harsh distortion and cannot be undone by lowering the level later.

The detector reads every sample of every channel and counts those at or above the threshold you choose. A single sample at full scale can be a legitimate peak that was normalised to 0 dBFS, so it is not called clipping on its own. A run of three or more consecutive samples at that level is a flat top, the typical signature of clipping, and is reported as a clip event with its time (minutes:seconds.milliseconds) and channel so you can jump to it in your editor. The first 50 events are listed; the counts include all of them.

Inter-sample overs

Between samples the reconstructed waveform can rise higher than any single sample. The detector oversamples the audio four times, as a true-peak meter per ITU-R BS.1770-5 does, and lists every place where the reconstructed waveform crosses the ceiling you set (−1 dBTP by default, the ceiling Apple Podcasts and EBU R 128 use). Overs flagged "between samples" are invisible to an ordinary sample-peak meter but can still clip in an MP3 or AAC encoder or in the playback device.

What to do with the results

• Clip events in the recording: lower the input gain and record again, or try a de-clipping tool on short, light cases.
• Overs but no clip events: put a true-peak limiter at the end of your chain, set to your delivery ceiling. The gain calculator shows how much limiting a loudness change needs.
• Neither: check the loudness with the LUFS meter before publishing.