Audio Analyzer
Measure sample peak, an interpolated true peak, RMS, a simple LUFS figure, and stereo correlation for an audio file.
🔒 Processed entirely in your browser — nothing you enter here is ever uploaded.
The result appears on this page after you run the tool. Nothing you enter is uploaded.
Result
About this tool
Upload an audio file and run the report. The page decodes it and walks every sample. Sample peak is the largest absolute sample on any channel. RMS is the square root of the mean of the squared samples, across channels. Both are printed in dBFS, which is 20 times the log10 of the amplitude, with a tiny floor so silence does not become negative infinity. A full-scale sine has peak 1, which is 0.00 dBFS, and RMS near 0.707, which is about −3.01 dBFS. Crest factor is peak dBFS minus RMS dBFS, so that sine is about 3.01 dB. DC offset is the average of the mid signal (left plus right, over 2).
The LUFS line is the RMS level in dBFS minus 0.691. It is not ITU-R BS.1770 integrated loudness: there is no K-weighting and no absolute or relative gate. True peak is the sample peak, plus the three straight-line points between each pair of neighboring samples. It is not a 4× oversampled true-peak meter. Correlation is the normalized sum of left times right, clamped to −1 through 1. A mono file, where right is a copy of left, reports correlation 1. Pan weight splits the balance of left RMS and right RMS into two percentages that add to 100.
Beside the numbers, a log-frequency spectrogram covers the mono mix from the start of the file through 60 seconds, with a 2048-point FFT and a hop of 512. A longer file is still measured in full for the numbers; only the picture stops at one minute. A second canvas draws the correlation bar, left and right RMS bars, and a pan marker. The report can be copied. The file is not uploaded.
Example. A full-scale sine has sample peak 1 (0 dBFS) and RMS about 0.707 (−3.01 dBFS), so the reported LUFS figure is about −3.70, which is that RMS level minus 0.691.
Limit. The LUFS line is an RMS offset, not gated ITU-R BS.1770 loudness, and true peak is the loudest sample or straight-line point between neighbors, not a 4× oversampled meter. The spectrogram covers only the first 60 seconds.
Is the LUFS number real integrated loudness?
No. The page subtracts 0.691 from the RMS level in dBFS and labels the result LUFS. It does not apply the K-weighting filter or the absolute and relative gates in ITU-R BS.1770, so it will not match a broadcast loudness meter on the same file.
What does the true-peak line measure?
It walks every sample and also the three evenly spaced straight-line points between each pair of neighbors, then keeps the largest absolute value. A real true-peak meter oversamples by at least four times. Use this figure as a hint that the waveform rose between samples, not as a compliance reading.
Why does the spectrogram stop at one minute?
The analyzer draws a log-frequency spectrogram of the mono mix with a 2048-point FFT and a hop of 512 samples, and it stops the picture at 60 seconds even when the file is longer. Duration, channel count, and the numeric report still cover the whole decoded file.