Audio Loudness Auditor
Check the perceived level of a local audio file before publishing it. The browser converts decoded audio to 48 kHz, runs the BS.1770-5 K-weighting filters, then averages overlapping 400 ms blocks through the absolute and relative gates. A four-phase FIR estimates intersample true peak. The chosen LUFS and dBTP values are comparison targets only, not a platform certification verdict. This is not an independently certified EBU Mode meter and does not measure loudness range (LRA).
Key features
- Measure gated integrated LUFS with the 48 kHz BS.1770-5 K-weighting coefficients and 400 ms blocks shifted by 100 ms.
- Estimate intersample true peak with the 48-tap, four-phase 48-to-192 kHz FIR example in BS.1770-5; also show actual sample peak.
- Inspect maximum 400 ms momentary and 3 s short-term loudness, each updated every 100 ms.
- Compare integrated level with a chosen reference and true peak with a chosen ceiling; inspect absolute and relative gate counts.
- Analyze in a cancelable browser worker and download a JSON audit including source SHA-256, settings and 400 ms block timeline.
How to use
- Choose a browser-decodable local audio file up to 8 MiB, or load the generated -18 dBFS stereo example.
- Set a LUFS reference (default -23) and a true-peak ceiling (default -1 dBTP).
- Run analysis and read integrated loudness, peak measurements, gate counts and the 100 ms timeline.
- Download the JSON manifest to reproduce the comparison; listen to the audio separately when making mastering decisions.
Use cases
- Check whether a spoken-word file is close to a chosen delivery reference.
- Find an intersample peak that is higher than the largest stored PCM sample.
- Compare two rendered versions using reproducible per-file JSON reports.
Frequently asked questions
How is integrated LUFS computed?
Each decoded channel passes the 48 kHz K-weighting shelf and high-pass filters from ITU-R BS.1770-5. The tool computes 400 ms block powers every 100 ms, rejects blocks at or below -70 LUFS, derives a second threshold 10 LU below the absolute-gated average, then energy-averages blocks above both thresholds. Mono and stereo channels have unit weight.
What does true peak measure that sample peak misses?
The decoded waveform can peak between its stored samples. The 48-tap, four-phase FIR example in BS.1770-5 reconstructs four positions per input sample at 48 kHz, giving a 192 kHz estimate. The result is reported in dBTP and is at least the stored sample peak. It is not a substitute for a certified measurement chain.
Is this an EBU R128 or EBU Mode certified meter?
No. The integrated and true-peak calculations follow the cited BS.1770-5 formulas and the -18 dBFS stereo alignment example is tested. This tool has not passed the complete official compliance suites, does not calculate LRA, and must not be treated as an EBU Mode certification.
Why can the same file read differently after upload?
The browser decodes the source codec and, when necessary, resamples it to 48 kHz so these published filter coefficients apply. Lossy codecs, decoder implementations, resampling and limits on short files can change the measured samples. Keep the source and JSON digest together when comparing runs.
What happens with silence and very short files?
If no block clears the absolute gate, integrated LUFS and any undefined peak are shown as a dash in the UI and null in JSON. A minimum of 0.4 seconds is required for one momentary block; a 3-second short-term maximum appears only if the file lasts at least 3 seconds.
Is the file uploaded or modified?
No. Audio is decoded and measured in browser memory. The JSON includes the source filename, byte size, SHA-256 and metrics but no PCM samples. The source is never overwritten.
Privacy
All audio decoding, hashing, filtering and peak analysis occur locally in the browser. No source audio or sample arrays are sent to this site; the optional JSON download contains measurement metadata and a source digest, not sound samples.
Comments & questions