Darwin's Cat

Free Online Clipping Detector & De-Clipper — Fix Clipping

    This free online clipping detector doesn’t just find the problem — it fixes it. Drop an audio file and it flags sample clipping (runs of consecutive full-scale samples, with timestamps), true-peak overs and DC offset per channel, then repairs the file in one click: de-clip the flat tops, remove DC, limit the true peak to −1.1 dBTP, and download a clean WAV. Most "clipping detector online" tools only detect; this one also repairs. It works on WAV, MP3, M4A, FLAC, OGG and AAC, runs 100% in your browser — no upload, no account — and keeps working offline after the page loads.

    The short answer

    A file is clipping when several samples in a row are pinned at digital full scale (0 dBFS). One lone full-scale sample is not clipping — clipping is a run. This tool flags a run of three or more consecutive samples at or above −0.1 dBFS, lists where each run is, and reports the longest one.

    True peak is a different measurement. It estimates the real analog peak between samples (the inter-sample peak), which can sit above the highest sample. A file can be true-peak-over with zero clipped samples — and it can have clipped samples while its true peak still reads below 0 dBTP. The tool shows both side by side and never mixes them up.

    If you just want a safe master: aim for no clipped sample runs and a true peak at or below −1 dBTP.

    Sample clipping vs true peak vs inter-sample overs

    These three terms get conflated constantly, which is why people ask "my DAW meter is fine, why does your tool say it clips?". They measure different things:

    Sample clipping is in the file already. When the signal was pushed past full scale during mixing, mastering or export, the samples were clamped to the maximum value. A row of identical max-value samples is the fingerprint of clipping. It is permanent and audible as a flat-topped waveform.

    True peak (inter-sample peak) is what a digital-to-analog converter actually reconstructs between the stored samples. Two samples can both sit below 0 dBFS while the smooth curve through them rises above it. That overshoot is an inter-sample over. Your DAW’s sample peak meter will not show it; a true-peak meter (this tool, and our LUFS meter) will.

    Lossy overs appear after encoding. AAC, OGG and MP3 reconstruct the waveform slightly differently, adding roughly 0.3–1 dB of extra peak. A file that is exactly at 0 dBTP before encoding can clip on playback after it. That is the practical reason for headroom.

    Why −1 dBTP

    Leaving about a decibel of true-peak headroom protects you from the lossy overs above. If you master to 0 dBTP, the streaming service’s encoder can push the playback peak past full scale and the listener hears clipping you never put there. −1 dBTP is the common safe ceiling; some engineers go to −1.5 dBTP for heavily limited material. For CD or files that are played as-is, true peak below 0 dBTP is the hard limit, ideally −0.3 dBTP.

    DC offset

    DC offset is a constant bias in the waveform — its average sits above or below zero instead of centering on it. It wastes headroom, can cause clicks at edits, and is usually a recording-chain or plugin artifact. This tool reports the per-channel offset in dBFS and flags anything above about −50 dBFS. A clean file should read far below that.

    How to read the result

    • Sample clipping: none — no full-scale runs were found. Good.

    • Sample clipping: N clipped runs — the file already contains clipping. Hit Fix it & download to reconstruct it here, or re-export from the source with more headroom. The list shows where each run starts, on which channel, and how long it is.

    • True peak within −1 dBTP — safe for lossy delivery.

    • True peak over −1 dBTP — below 0 dBFS but tight; lossy codecs may push it over. Lower the limiter ceiling.

    • True peak clipping (≥ 0 dBTP) — the reconstructed peak is at or above full scale. Reduce gain or the limiter ceiling.

    Use Copy report for a plain-text summary, or Copy CSV to paste every clip location into a spreadsheet.

    Fix it — one click

    When clipping or a true-peak over is found, the Fix it & download button repairs the file right here, in your browser. It only ever fixes measurable technical faults, and the result tells you exactly which steps ran:

    1. De-clip — the flat-topped runs we located are reconstructed: each clipped run is replaced with a smooth curve through the good samples on either side, restoring the rounded peak the clamp destroyed. This is the real fix for baked-in sample clipping — not just turning it down.

    2. DC offset removal — if the waveform sits off-centre, the constant bias is subtracted so it centres on zero again.

    3. Mono fix — if the file is stereo but one channel is silent (a mono source recorded into one side), the active channel is copied to both.

    4. Safe peak — the steps above can overshoot past full scale, so the true (inter-sample) peak is brought to −1.1 dBTP last. The result reads genuinely clean on a re-check, not merely quieter.

    You get a fresh 24-bit WAV (nothing is uploaded) and a second player to A/B the repaired audio against the original.

    Experimental extras (off by default). Optional, clearly-labeled treatments you can tick before fixing: Voice cleanup (an 80 Hz–8 kHz band-pass — for speech only; it makes music dull), Trim silence (crop dead air at the head and tail), and Repair clicks (interpolate isolated digital spikes). Best-effort, and they stay off unless you ask.

    Not every broken file can be fully repaired — but we do everything that can honestly be done. Long clipped runs lost more information, so they are only partially recoverable: we round the top back, but samples the clamp threw away cannot be invented. For heavy, sustained clipping the only true fix is a new export from the source with more headroom. We also deliberately do not do EQ, compression, loudness or noise removal here — that is tone-shaping, not repair: the mastering tool runs a full true-peak limiter against each platform, while the LUFS meter checks loudness and true peak.

    Under the hood

    Everything runs in your browser (Web Audio decode + a native WAV codec) — no upload, no server. The techniques we use:

    • Clip-run reconstruction — cubic-Hermite interpolation across each flat-topped run, with an adjustable slope window for the boundary estimate.

    • True-peak measurement & limiting — ITU-R BS.1770 4× oversampling; the repaired file is brought to −1.1 dBTP.

    • DC-offset removal — per-channel mean subtraction.

    • Mono / one-sided repair — detect a silent channel and restore both.

    • Sample-clipping detection — runs of consecutive at/near-full-scale samples, with per-channel timestamps.

    • Experimental treatments — RBJ biquad band-pass (voice cleanup), silence trim with micro-fades, and isolated-click interpolation.

    This page is part of a small set of free, no-upload audio tools — everything runs in your browser and your file never leaves your device.

    Frequently Asked Questions

    How do I check if my audio is clipping?

    Drop the file here — nothing is uploaded. The detector flags sample clipping (runs of consecutive full-scale samples) with timestamps, estimates the true peak (inter-sample), and measures DC offset, then gives an honest pass/fail.

    Can clipped audio be repaired after recording?

    Partly. The one-click fix de-clips by reconstructing the flat-topped runs (cubic-Hermite interpolation through the good neighbouring samples), then limits the true peak to −1.1 dBTP and exports a clean WAV. Short clips recover well; long, heavy clipping lost information and is only partially recoverable — for that, re-export from the source with more headroom.

    What’s the difference between sample clipping and true-peak (inter-sample) clipping?

    Sample clipping is already baked into the file — samples clamped flat at digital full scale (0 dBFS). True peak is the analog peak reconstructed between samples; it can exceed 0 dBTP even when no single sample reads full-scale. A file can be true-peak-over with zero clipped samples, and clipped with a true peak below 0 dBTP — the tool reports both separately.

    Why does audio clip after converting to MP3, AAC or streaming?

    Lossy encoders (MP3/AAC/OGG) reconstruct the waveform slightly differently and add roughly 0.3–1 dB of inter-sample peak. A file at exactly 0 dBTP before encoding can clip on playback afterwards. That's why you leave headroom and limit to −1.1 dBTP.

    How do I remove DC offset from a WAV file?

    The fix subtracts the per-channel mean so the waveform re-centres on zero. DC offset wastes headroom and can cause clicks at edits; it's reported in dBFS and removed automatically when present — all in your browser.

    Why limit the true peak to −1 dBTP?

    About a decibel of true-peak headroom protects against the inter-sample overs lossy codecs add. Master to 0 dBTP and a streaming encoder can push playback past full scale; −1 dBTP (we use −1.1 for margin) keeps it safe. For CD or as-is files, stay below 0 dBTP.

    Is my file uploaded? Is it private?

    No upload. Everything — detection, de-clipping and the WAV export — runs locally in your browser; the file never leaves your device, and there's no account or sign-up.

    Will the repaired audio sound exactly like the original?

    Not exactly — de-clipping is an honest estimate, not magic. It reconstructs the rounded peaks the clamp destroyed and removes the worst distortion, but samples that were thrown away can't be invented. Lightly clipped material is often indistinguishable; heavily clipped material improves but isn't perfect.

    Comments

    Be the first to leave a comment.

    By submitting you agree to our privacy notice.