# Detect Clipping in WAV Audio
Paste base64 or hex WAV bytes and detect clipped full-scale samples, consecutive clipping runs, timestamps, and worst distorted regions. Runs locally in your browser.
## Run it
- **CLI:** `gizza tool clipping-detector "UklGRjQAAABXQVZFZm10IBAAAAABAAEACAAAABAAAAACABAAZGF0YRAAAADNDJkZ/3//f/9/ZibNDAAA"`
- **Web:** https://gizza.ai/tools/clipping-detector/
- **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/clipping-detector/tool.json
## Inputs
- `input` — WAV bytes (base64 or hex) _(field)_
- `input_format` — Input encoding _(field)_
- `output` — Output _(field)_
- `threshold` — Clipping threshold (fraction of full scale) _(field)_
- `min_run` — Minimum run length (frames) _(field)_
- `gap` — Bridge gap (frames) _(field)_
- `top_regions` — Worst regions to list _(field)_
## Output
- Clipping report (text)
## Query parameters
Open the tool pre-filled and auto-run via URL:
- `input` — WAV bytes (base64 or hex)
- `input_format` — Input encoding
- `output` — Output
- `threshold` — Clipping threshold (fraction of full scale)
- `min_run` — Minimum run length (frames)
- `gap` — Bridge gap (frames)
- `top_regions` — Worst regions to list
Example: `https://gizza.ai/tools/clipping-detector/?input=UklGRjQAAABXQVZFZm10IBAAAAABAAEACAAAABAAAAACABAAZGF0YRAAAADNDJkZ%2F3%2F%2Ff%2F9%2FZibNDAAA&input_format=base64&output=report&threshold=0.99&min_run=1&gap=0&top_regions=5`
---
## About this tool
Digital clipping happens when audio samples hit full scale and flatten against the ceiling. A waveform can look mostly fine while a few short runs of full-scale samples create harsh clicks, crunch, or distortion. This tool decodes a short uncompressed WAV clip and reports exactly where those clipped samples occur.
Paste WAV bytes as **base64** (default) or **hex**. The detector supports uncompressed RIFF/WAVE PCM (8/16/24/32-bit integer) and IEEE float WAV (32/64-bit). It counts individual clipped samples, sample-frames with any clipped channel, the longest consecutive run, the overall peak in dBFS, and the worst regions with start/end timestamps.
### Worked example
The sample preset contains an 8 Hz mono WAV with three consecutive full-scale samples. Run it with the default `0.99` threshold and you will see one clipped region from `0:00.250` to `0:00.625`, `3 of 8` clipped samples, and a longest run of `3` frames. Set **Output** to `json` when you want structured counts and region objects for another script.
### Controls
- **Threshold** is the absolute sample magnitude that counts as clipped. `0.99` catches true 0 dBFS peaks with a small guard band; `1.0` only catches exact full-scale samples.
- **Minimum run length** filters out single-sample spikes. Audacity's classic clipping check uses `3` consecutive clipped samples.
- **Bridge gap** merges runs separated by a few clean frames so one distorted burst is not split into many tiny regions.
- **Worst regions** limits how many regions are listed, ranked by length and then by peak.
### Limits and edge cases
This is a sample scanner, not a decoder for every audio format. Convert MP3, AAC/M4A, FLAC, Ogg/Opus, A-law, or mu-law audio to uncompressed WAV first, then paste those bytes here. Very large audio should be clipped to the section you want to inspect before base64/hex encoding; the page is designed for short diagnostic clips. Stereo and multichannel files are scanned frame-by-frame: one clipped channel makes the frame count as clipped, while the sample count still counts every clipped channel sample.
## FAQ
What does "clipping" mean?
Clipping means a sample reached the maximum representable amplitude, so the waveform is flattened at the top or bottom instead of following the original shape. A few isolated full-scale samples may be harmless, but consecutive runs often mean audible distortion.
Why do I need WAV bytes instead of uploading an MP3?
This pure WebAssembly tool inspects decoded PCM samples directly. MP3/AAC/FLAC/Ogg need a codec decode step first, so they are rejected with a clear message rather than guessed at. Convert the file or the suspect section to uncompressed WAV, then paste its base64 or hex bytes.
Which threshold should I use?
Use `0.99` for a practical scan: it catches samples that are effectively full-scale even if a codec or export step nudged them slightly below 1.0. Use `1.0` when you only want mathematically exact full-scale samples. Lower values such as `0.95` are useful for finding near-clipping headroom problems.
What is a clipped region?
A region is a consecutive run of sample-frames where at least one channel is clipped. `min_run` decides how long the run must be before it is listed, and `gap` can bridge a few clean frames between clipped frames when one distorted burst is broken into fragments.
## Related tools
- [Speech audio quality checker](https://gizza.ai/tools/speech-audio-quality-checker/): Paste a short WAV clip as base64 or hex and check ASR readiness: sample rate, channels, SNR estimate, clipping, level, and verdict.
- [Video audio RMS timeline](https://gizza.ai/tools/video-audio-rms-timeline/): Extract windowed RMS and peak audio levels from a video's audio track into a CSV or JSON time series. Set the window, hop, and dBFS/linear unit.
- [Base Decoder](https://gizza.ai/tools/base-decoder/): Auto-detect and decode Base16, Base32, Base45, Base58, Base64, and Base85 text, including nested layers, locally in your browser.
- [Decode/Transform Recipe Pipeline](https://gizza.ai/tools/cyberchef-pipeline/): Chain byte-level decode/transform steps — Base64, hex, URL, gzip/zlib, XOR, ROT13 — into one recipe, applied top to bottom. Free, private, runs entirely in your browser.
- [Downsample Time-Series (LTTB)](https://gizza.ai/tools/downsample-timeseries/): Reduce a large time-series to a target point count with LTTB, min/max, M4, or every-nth downsampling — shape preserved, in your browser, no upload.