How to analyze an audio file
Choose one audio file with browse files or drop it into the upload area. The picker accepts audio/*; the drop zone lists MP3, WAV, FLAC, AAC, OGG, and more. It accepts one file at a time.
- Select or drop the file.
- Wait for decoding and analysis to finish. Processing runs locally in your browser and does not change the source file.
- Use the audio transport, then click the waveform or spectrogram to seek to a position.
- Click a clipping or silence event to seek to its listed timestamp.
Review the analysis results
The waveform shows amplitude over time. The spectrogram shows frequency content over time; brighter areas indicate more energy. Both views can be saved with their Export PNG controls.
Review BPM and key estimates with their confidence values and method notes. The report also includes integrated loudness (LUFS), RMS, sample peak, true peak (dBTP), crest factor, loudness range, clipping events, silent intervals, and timestamps.
Stereo phase correlation is shown only when the decoded audio has at least two channels. The phase panel is hidden for mono files. Its scale runs from +1 for identical channels to -1 for fully inverted channels.
Use measurements as estimates
- BPM uses onset-strength autocorrelation. It searches 40-220 BPM, then normalizes the displayed result into the 70-180 BPM range. Syncopated or sparse material can produce a half- or double-tempo result, so check the confidence value.
- Key is estimated from an average chroma vector matched against major and minor key profiles. Treat it as an estimate, especially for atonal, noise-heavy, heavily percussive, or modulating material.
- Integrated loudness (LUFS) uses gated K-weighting based on ITU-R BS.1770. It is an independent implementation, not a certified loudness meter.
- RMS is the full-signal level in dBFS. Sample peak is the loudest sample, while True peak (dBTP) uses 4x oversampling to detect inter-sample peaks.
- Clipping events mark runs at or above -0.3 dBFS. Silent intervals mark levels below -50 dBFS for at least 300 ms.
Export the analysis
Use Export PNG in either visual panel, Export JSON report for structured results, or Export text report for a readable summary. The text report includes at most 50 clipping events and 50 silence events per category.
File handling, processing and limits
Analysis is local and read-only: the tool does not write to, re-encode, trim, or normalize the source file. If the browser cannot decode the selected file, it shows an alert beginning Could not decode this audio file. The visible clipping and silence lists render at most 200 events each, even when the report detects more. Phase correlation is unavailable for mono input.
Frequently asked questions
What files can I select?
The picker uses audio/*. The drop zone displays MP3, WAV, FLAC, AAC, OGG, and more, with one file supported at a time.
What happens if the file cannot be decoded?
The tool shows an alert beginning Could not decode this audio file. No analysis results are produced for that file.
How accurate is the BPM estimate?
BPM is an estimate from a 40-220 BPM search with displayed-result normalization to 70-180 BPM. Syncopated or sparse material can produce half- or double-tempo results, so use the confidence value and treat the result as a starting point.
Why is the phase panel missing?
Phase correlation requires at least two channels. The panel is hidden when the input is mono.
What can I export?
You can export the waveform or spectrogram as PNG, or use Export JSON report and Export text report. The visible event lists show up to 200 clipping and 200 silence events; the text report includes up to 50 of each category.
Is the LUFS reading broadcast-certified?
No. It uses an independent browser implementation of the gated K-weighted method described in ITU-R BS.1770, not a certified loudness meter.
Why does true peak differ from sample peak?
True peak uses 4x oversampling to estimate peaks between samples. Sample peak only measures the stored sample values.