Free tool
Spectrum analyzer
Drop in a track and see its average spectrum — the whole file's tonal balance in third-octave bands, with the tilt and centroid figures that say in numbers what the shape says in outline.
Drop an audio file, or choose one. It is read on this device and never sent anywhere.
If the balance is off
A spectrum shows the shape; mastering decides what to do with it
One contextual next step, offered after your result — never in front of it.
A tilted or lumpy average spectrum is a description, not a diagnosis — the same curve can be a defect in one genre and the sound of another. Euphona’s AI mastering analyses the mix against reference curves for the material, applies corrective equalisation with the reasons attached, and plays the result back level-matched so you judge the balance and not the volume. The free plan runs it on your own track.
The numbers
How to read an average spectrum
What the bands, the tilt and the centroid each tell you — and the things a spectrum can't say.
An average spectrum shows where a track’s energy sits across frequency, summed over the whole file. Unlike the dancing display in a real-time analyzer, it's a stable measurement you can compare between mixes — which is what makes it useful for judging tonal balance rather than for watching a signal move.
- Third-octave bands
- The hearing-shaped summary the audio world has used since analogue analyzers: each band is a third of an octave wide, so the display gives low, mid and high frequencies the space the ear gives them. Raw FFT bins would spend half the plot above 10 kHz, where a twelfth of the hearing range lives.
- Tilt
- The overall slope of the spectrum, in dB per octave. Zero reads like pink noise — equal energy per octave, the reference most mastering analyzers are calibrated around. Positive is brighter than that, negative darker. Full mixes commonly sit somewhat negative; how far is a genre question, not a rule.
- Centroid
- The frequency the energy balances at — the spectrum’s centre of mass. It tracks perceived brightness in one number: two mixes with the same tilt can still differ here when one carries its energy in different places.
- What it can't say
- A spectrum sums everything, so it can't tell an equalisation problem from an arrangement one — a 300 Hz bump might be mud, or it might be the cello section. It also says nothing about time: masking, dynamics and transients are invisible here, which is why a spectrum is a check on your ears and not a substitute.
Verification
How to check this tool
The measurement is a Welch-averaged FFT folded into third-octave bands — commodity signal processing, written from the textbook, checked on every build against constructed signals whose answers are known: a sine lands in its own band at 0 dB, equal-energy-per-band material reads flat, white noise reads +3 dB per octave. Compare the band readings against any analyzer you already trust; they should agree within a band’s width.