Treat each file as its own programme
Each selected file receives an independent integrated measurement and its own processing decision. The batch does not concatenate files or infer that separate tracks should share one album-level gate. Choose excerpts and boundaries consistently if you want meaningful comparisons across a set of recordings.
The batch CSV and JSON preserve source names, loudness, true peak and processing settings. A long silence before or after content can be excluded by loudness gating while still remaining in the exported audio. The tool does not trim duration, remove breaths or synchronize programme boundaries for you.
Interpret range and history together
Loudness range estimates the spread of gated three-second loudness windows using lower and upper percentiles. It is different from peak-to-average ratio and from the numerical difference between the loudest and quietest sample. The history plot shows where changes occur, while LRA summarizes their distribution.
Material shorter than 60 seconds is marked provisional for LRA. A short dramatic excerpt may not represent the complete programme, and a near-silent clip may provide too few valid windows. An unavailable value stays unavailable rather than being presented as a meaningful zero.
Keep the comparison bounded and repeatable
Large decoded audio consumes much more memory than a compressed file. The workbench limits combined file size and the total number of decoded channel samples, processes up to three files, and runs DSP in a cancellable local worker. Split a larger set into small batches while keeping the same target and ceiling.
The waveform and audio players use the same decoded source and verified export that were measured. Playback starts only after you press Play, at a reduced initial volume. Download the WAV and report before closing the tab, then reopen a representative result in your normal audio editor to confirm the channel layout and intended listening effect.