Analyze audio
What Analyze audio actually does
Every consumer platform turns a track up or down at playback until its measured loudness matches a fixed level, so the number that decides how loud a listener hears a recording is calculated after upload and is almost never shown to whoever made it. That number is integrated loudness, expressed in LUFS, and it comes from an algorithm standardised as ITU-R BS.1770 that weights the signal for human hearing and ignores the quiet passages instead of averaging them in. A peak meter cannot answer it: two files can peak at the identical value and sit several LUFS apart, and the quieter one will be turned up while the louder one is turned down. This page measures the loudness figure, the loudness range that says how far the loud and quiet passages sit apart, and the true peak, which is measured between samples rather than only at them and is therefore the one that predicts whether a lossy encoder will clip. Each figure is stated against the target you pick, in text, with the difference in LU. Beside them the same file is drawn twice: a spectrogram on a logarithmic frequency scale, so the musical range is spread evenly rather than squeezed into the bottom of the picture, and a waveform that plots the loudest sample in each column and separates the channels, so a short transient and a lopsided mix are both visible instead of being averaged away.
How to use it
- Choose one audio file. It is decoded in this tab and its length, sample rate and channel count are read before anything is measured.
- Pick the level your release is aimed at, and the way the pictures should be drawn: logarithmic or linear frequency, peak or average waveform sampling, channels together or apart, and the image size.
- Analyze. Levels are measured in a pass of their own, ahead of the drawing, so a picture that fails to render still leaves the numbers on the page.
- Read the verdict, the loudness table and the per-channel levels, then download either picture or the whole measurement as JSON.
Useful for
- Find out whether a master will be turned down at playback, and by how much, before it is uploaded anywhere.
- Check a true peak against the ceiling a lossy encoder needs, which a sample peak meter will read as safe when it is not.
- See whether one channel is quieter than the other, or whether a recording has a noise floor or a constant offset, without opening a full editor.
- Look at what actually occupies the top of the spectrum in a recording, on an axis where the musical range is legible.
Limits worth knowing
- Tempo in beats per minute and musical key are not measured. Both need a library this browser build leaves out, and a guessed answer is worse than no answer.
- The tolerance band used to call a level at target is this tool's own stated band, not a quoted requirement from any standard or platform.
- Loudness is measured over the whole file. A short measurement window and a moving meter are a different job from a delivery check.
- The spectrogram's requested size is the plot area, and the legend is drawn outside it, so the produced picture is larger than the size requested. The size actually produced is read back from the image and reported with it.
Questions people ask
undefined
undefined
undefined
undefined
undefined
undefined
undefined
undefined