Generate & process audio
Audio analyzer
Load audio or try the piano. Choose Analyze to see loudness, peaks and stereo readings.
Looking for a different task? Measure loudness · Review listening points and check mono.
1 · Choose audio to analyze
MP3 or WAV · up to 10 min and 128 MB · mono or stereo.
Tap the waveform to seek.
Minutes:seconds.milliseconds or seconds.
Volume affects playback only.
Analysis runs on your device and does not modify the original file.
2 · Results
Measurements appear here after analysis.
How to review your audio
- Choose a file or Try a piano example. Play it before checking the numbers.
- Keep Whole track or choose an excerpt. Start and End set the measured interval; the white cursor only moves playback.
- Choose Analyze track or Analyze excerpt. Review Levels, Frequency and Stereo, then export the report if you need to keep it.
What each measurement means
- Load a file your browser can decode, up to 10 minutes and 128 MB, mono or stereo. It is read and analyzed on your device. Keep the original.
- Integrated LUFS summarizes loudness through weighting and gating; it does not measure artistic quality. LRA describes variation and needs sufficient temporal context. Sample peak measures maximum amplitude and estimated true peak accounts for peaks between samples. Integer PCM and floating point have different headroom: touching 0 dBFS alone does not prove clipping. RMS describes effective amplitude and crest factor compares peak and RMS; a low value does not establish that a compressor was used.
- The loudness over time shows how the loudness changes along the file. The cyan line is the short-term loudness (measured in 3-second windows, every tenth of a second) and the amber one is the momentary loudness (400-millisecond windows); the light horizontal line is the integrated loudness of the analyzed interval, which ignores very quiet stretches and may therefore not match the average you see in the curve. Each value is placed at the moment its window closes, so the momentary one starts at 0.4 seconds and the short-term one at 3: in a shorter file there is no curve, or no short-term one. The maximum of each appears in the Level group. Move the cursor, or use the left and right arrow keys (with Shift, 10 seconds at a time), to read the values.
- The full-scale sample counter uses a 0.9999 threshold and also counts runs of three or more. It locates areas to listen to and review; it does not by itself prove audible clipping. DC offset is the signal average and may consume headroom.
- The analyzer tries to read format, sample rate and depth from file bytes. Browser decoding can resample to the audio context’s rate, which need not be your sound device’s rate. If HertzKit cannot preserve or identify the original rate, it shows the rate used for analysis. MP3 and AAC are not described by a PCM bit depth; Ogg is a container for different codecs. The effective bits reading examines sample patterns in supported formats: it does not certify origin, quality, noise or converter resolution. Gain, processing or dither can change those patterns; they are not calculated for FLAC.
- The analyzer does not report whether a file was dithered or what kind of dither, and that is on purpose: in a finished file, in general, it cannot be known. The end of a fade-out with triangular dither is practically indistinguishable from a recording with analog noise, and there is only something to look at in very low-level passages. Reporting it would be making up data.
- In stereo, correlation describes the relationship between channels: +1 may mean proportional signals with the same polarity, 0 means little linear relationship and −1 means opposite polarities. Complete cancellation also requires equal amplitudes. Balance compares left/right energy and width compares side and mid energy. Listen to the mono sum before correcting a number.
- The average spectrum shows how much energy there is at each frequency, averaged over the analyzed interval, in calibrated dBFS (a full-scale sine wave reads 0 dBFS). The cyan line is the power average and the amber one is the maximum each frequency reached at any moment. Move the cursor, or use the left and right arrow keys (with Shift, one octave at a time), to read the frequency, the note and the level. A sine wave that falls between two of the frequencies the FFT computes reads up to 1.4 dB lower than it really is.
- The ten bands, from 31.5 Hz to 16 kHz, group estimated spectral energy across windows of the mono average. They need not sum exactly to the channels’ global RMS. The ends have resolution and coverage limits; partial bands are flagged. No distribution makes every mix correct.
- The goniometer plots stereo sample pairs: equal signals form a vertical line and equal signals with opposite polarity form a horizontal one. Use it to help check width and the mono sum; it does not grade a mix or replace listening to changing passages.
- The spectrogram shows how the frequency content changes over time: the horizontal axis is time, the vertical axis is frequency (low at the bottom, high at the top), and brightness shows how much energy is at each point. The vertical axis is logarithmic, not linear — so an octave always takes up the same space whether it's in the lows or the highs, and bass doesn't get squashed against the floor. Move the cursor to read the time, frequency and level of any point.
- The dominant frequency comes from the highest peak of the average spectrum above 20 Hz, refined between neighboring frequencies, with its note and cents deviation. It is still a rough indicator: in a full mix it will almost always reflect the strongest element (often the kick or the bass), not necessarily what your ear perceives as the lead. An amber line on the spectrogram marks where it falls.
- Spectra, bands and dominant frequency use (L + R) / 2. Opposing content can be reduced or cancel there while channel energy remains. RMS and stereo data use their own calculations. The analyzer guide explains how to compare these readings.
- Analyzing a file doesn't change who holds the rights to that audio: if it isn't yours, make sure you have permission before using it. This is a general reminder, not legal advice.
Files and progress
Up to 10 minutes and 128 MB per file, mono or stereo. Both limits apply. Supported formats and available memory depend on your browser and device; if you have trouble, try a shorter excerpt.
Analysis runs in several stages. Use “Cancel” while it is running. After canceling, choose the analysis button to retry with the same file. Selecting another file discards the previous analysis.
Try a practical example
Choose Try a piano example, listen to the opening and analyze the whole track. Then choose an excerpt from 0 to 3 seconds and analyze it again. Compare its loudness curve with the full track. Both results describe different intervals; neither is a grade for the performance.
Frequently asked questions
Does analyzing change my audio?
No. Playback and analysis do not rewrite your original file or create a converted download. Your browser decodes audio to play and measure it; playback volume changes only what you hear. What changes when exporting from other tools?
Does analysis tell me whether my mix sounds good?
No. It describes levels and energy distribution without knowing your intent, genre or arrangement. Measurements help locate details and compare changes; they are not an artistic score or quality certification.
Do RMS, LUFS and peak measure the same thing?
No. Peak describes maxima; RMS summarizes effective amplitude; LUFS uses frequency weighting and gates to estimate loudness. No single measurement replaces listening. True peak is estimated: check demanding deliveries with an appropriate reference.