L/R RMS balance
20 log10(L_RMS / R_RMS), in dB, for exactly two channels. Positive means higher left RMS amplitude, negative means higher right RMS, and zero means equal RMS. Complete unwindowed frames use 4096 samples and hop 1024 by default; times mark frame centers relative to the analyzed audio. Each channel must meet the configured RMS floor (default -80 dBFS); otherwise the ratio is undefined with a channel-specific reason. The floor is an analysis limit, not a quality threshold. Mono and multichannel inputs are not applicable. Statistics exclude undefined frames. This does not establish pan position, perceived image position, loudness, a defect, or whether asymmetry is intentional.
LUFS
Integrated LUFS measures hearing-weighted loudness over the analyzed audio, including only the selected range when one is used. It gives delivery context, not a quality judgment.
Short-term LUFS
Short-term LUFS is a time-varying K-weighted loudness measurement using the configured window (3 seconds by default). Points mark window ends: each value integrates the preceding block of K-weighted audio; a final block may be truncated. Filter state also depends on earlier samples. It is distinct from RMS and integrated loudness.
Approximate true peak
Approximate true peak estimates reconstructed peaks between samples. It is not a standards-grade true-peak measurement. Values above 0 dBTP can matter for playback, conversion, or encoding headroom.
Sample peak
Sample peak is the largest absolute decoded sample amplitude across all channels in the analyzed audio, expressed in dBFS. It is not loudness or a reconstructed intersample peak.
RMS
RMS is root-mean-square signal amplitude, expressed in dBFS. The headline RMS pools samples across all channels; the RMS timeline uses their arithmetic-average mono signal, where opposing channels can cancel. Neither is hearing-weighted loudness.
Crest factor
Crest factor is sample-peak-to-RMS contrast in dB. Both the headline and each crest frame use all channels, unlike the mono RMS timeline. It describes measured peak contrast, not punch, quality, or dynamic range.
PLR
PLR is approximate true peak minus integrated loudness. It describes the measured distance between those two values; it does not by itself identify compression, transient quality, or a delivery problem.
Clipping / near-clipping
Clipping counts samples at the configured ceiling; near-clipping counts samples close to it. Use the waveform and histogram to inspect where these samples occur.
Stereo correlation
Stereo correlation is Pearson r between left and right, from -1 to +1. +1 means matching variation, not necessarily equal levels; 0 means little linear correlation; negative values mean opposing variation. It does not directly measure perceived width or establish a phase defect.
Side/mid ratio
Mid = (L + R) / 2 and side = (L - R) / 2. Their RMS ratio is expressed as side/mid dB: negative means mid-dominant, zero means equal RMS, and positive means side-dominant. Undefined ratios are not zero. This does not directly measure perceived width or isolate sources.
Spectral centroid
Spectral centroid is the magnitude-weighted mean frequency of each mono spectrum, in Hz. Higher values mean more weight at higher frequencies, not higher note pitch or an EQ recommendation.
Rolloff
Spectral rolloff is the frequency in Hz below which 85% or 95% of summed spectral magnitude lies in a mono frame. It is not a filter cutoff or an audible-bandwidth estimate.
Spectral bandwidth
Spectral bandwidth is the magnitude-weighted root-mean-square frequency spread around the centroid, in Hz. Higher values mean more distributed spectral magnitude, not greater musical richness.
Average spectrum
The Welch average spectrum describes the arithmetic-average mono signal over the analyzed audio. For measurable audio, 0 relative dB marks the strongest averaged bin at or above 20 Hz. Values describe shape within this analyzed view, not calibrated dBFS or absolute levels across files.
Relative mean band power
This mono-downmix view averages spectral power per included FFT bin in each broad frequency band, then normalizes values within the analyzed view. Between-view differences do not establish absolute band-level changes. It is not integrated total band energy and does not indicate absolute dBFS level.
Onset density
Onset density is a smoothed onset-strength activity measure, not a count of events per second. Summary values retain the raw scale; each plotted curve is independently normalized to a maximum of 1 (or zero for no activity). It is not punch, groove quality, drum hits per second, or mix quality.
Chroma CQT
Chroma CQT shows normalized pitch-class energy, folded across octaves. Each nonzero frame is normalized to a maximum of 1. It is not exact note pitch (F0), tuning accuracy, key detection, or chord detection; values are not calibrated across unrelated songs.
Relative dB
Relative dB plots show shape within this track. They are useful for shape within this track; not comparable to dBFS values from meters or other songs.