No audio loaded. Open the Analyzer, load a file or enable the microphone, then return here for live context about your specific sound.
Learn - Overview
What is Soniform measuring?
Every sound is a pressure wave. What makes a violin sound different from a trumpet is the hidden mathematical structure inside that wave. Soniform decomposes sound in real time using the Fourier Transform and displays that structure across seven measurements and four visualizations.
Each metric in the bottom bar is recomputed every animation frame from your audio. Together they form a spectral fingerprint of the sound.
The seven metrics at a glance
Fundamental
The lowest, strongest pitch frequency. What a musician calls the note.
Detected Note
Nearest musical note name (A, C-sharp, etc.) plus cents sharp or flat.
Spectral Centroid
Centre of mass of the spectrum. High value means bright. Low value means warm.
RMS Energy
Overall loudness of the signal in decibels.
Zero Crossing Rate
How many times the waveform crosses zero per frame. A proxy for tonality versus noise.
Harmonic Complexity
Score from 0 to 100 measuring how evenly energy spreads across the overtone series.
Spectral Rolloff
Frequency below which 85% of the spectrum energy lives.
Use the sidebar to explore any metric in depth. If audio is loaded, each section shows your live reading and explains what it means for the sound you are hearing right now.
Learn - Fundamental Frequency
Fundamental Frequency
When a string or air column vibrates, its base rate of oscillation is the fundamental frequency (f0), the pitch your ear identifies.
f0 = 1 / T where T is the period of one full cycle in seconds
Soniform detects f0 using autocorrelation: it slides a copy of the waveform across itself and finds the time-lag where it matches best. That lag equals one period, giving the fundamental.
--
Your current fundamentalThe dominant pitch frequency detected in real time.
Middle C on a piano is 261.6 Hz. Human speech sits 85 to 255 Hz. A piccolo can reach 4 kHz. High number means high pitch; low number means bass.
autocorrelationpitch detectionperiodicity
Learn - Detected Note
Detected Note
Western music divides the octave into 12 equal semitones. Soniform maps the detected fundamental to the nearest note using:
semitones = 12 times log2(f0 / 440)
440 Hz is A4, the international tuning reference. Cents are hundredths of a semitone. A reading of 0 cents means perfectly in tune. A reading of 50 cents means halfway between two notes.
--
Nearest musical noteIncludes octave number and cents deviation from perfect tuning.
Singers and instrumentalists target 0 cents. Vibrato oscillates plus or minus 20 to 50 cents rhythmically.
equal temperamentA440cents
Learn - Spectral Centroid
Spectral Centroid
Imagine the frequency spectrum as a see-saw. Each frequency bin has a weight equal to its amplitude. The centroid is where that see-saw balances, the centre of mass of the spectrum.
centroid = Sum(fi times |X(fi)|) / Sum(|X(fi)|)
It is the single best predictor of perceived brightness. High centroid means tinny, bright, metallic. Low centroid means warm, dark, bass-heavy.
--
Spectral centre of massCorrelates directly with how bright or dark the sound is perceived.
A soft cello might show 600 Hz. A cymbal crash might show 8000 Hz.
timbrebrightnessMIR
Learn - RMS Energy
RMS Energy
Root Mean Square measures effective signal power. It tracks perceived loudness far better than peak level.
RMS = sqrt( (1/N) Sum x[n]^2 ) converted to dB = 20 times log10(RMS)
Silence is near -90 dB. A quiet room is around -60 dB. A speaking voice sits around -30 to -15 dB. Loud music approaches 0 dB. Above 0 dB indicates clipping.
--
Signal power in decibelsLoud is near 0 dB. Silence is near -90 dB.
loudnessdecibelssignal power
Learn - Zero Crossing Rate
Zero Crossing Rate
Every time the waveform crosses the zero axis that is a zero crossing. ZCR counts how many happen per analysis frame.
ZCR = Sum |sign(x[n]) - sign(x[n-1])| / 2
A pure 440 Hz sine wave crosses roughly 41 times per 2048-sample frame. Noisy broadband sounds cross hundreds of times. Clean tonal sounds cross far fewer times.
--
Zero crossings per frameLow means tonal or pitched. High means noisy or percussive.
noisinesstonalitytransients
Learn - Harmonic Complexity Score
Harmonic Complexity Score
A pitched sound distributes its energy across a harmonic series, the fundamental and its integer multiples. How that energy is spread tells us about timbre.
HCS = -Sum pn times log2(pn) / log2(N) times 100
This is the Shannon entropy of the overtone distribution, normalised to 100. A score of 0 means all energy is in one harmonic (pure sine wave). A score of 100 means energy is perfectly equally spread across all harmonics.
--
Overtone entropy (0 to 100)Flute is around 20 to 35. Guitar is around 50 to 70. Full orchestra is 80 and above.
This is Soniform's original metric. Simple flutes concentrate energy at the fundamental. Violins and brass spread it across many overtones producing richer scores.
entropyovertonestimbreoriginal metric
Learn - Spectral Rolloff
Spectral Rolloff
Rolloff is the frequency below which 85% of total spectral energy lives. Think of it as a brightness threshold.
Find f* such that: Sum over f less than or equal to f* of |X(f)|^2 = 0.85 times total energy
High rolloff means significant high-frequency content such as cymbals or consonants. Low rolloff means bass or kick drum is dominant.
--
85% energy thresholdBass-heavy sounds have low rolloff. Bright or noisy sounds have high rolloff.
brightness85th percentileenergy distribution
Learn - FFT and Spectrum
The FFT Spectrum
The Fast Fourier Transform takes a snapshot of N time-domain samples and decomposes it into constituent sine waves at every frequency from 20 Hz to the Nyquist limit.
X(f) = Sum from n=0 to N-1 of x[n] times e^(-i 2 pi f n / N)
Soniform plots magnitudes on a logarithmic frequency axis so octaves appear equally spaced, matching how human hearing works.
FFT Size
Larger FFT means more frequency bins and sharper spectral detail, but each frame covers more time so fast transients blur. Smaller FFT means faster response but coarser resolution. 2048 is the default balance.
Windowing
The FFT assumes the signal repeats periodically. A hard edge causes spectral leakage. Windowing tapers the frame to zero at both edges. Hann is a good general default. Blackman reduces leakage further. Rectangular gives maximum resolution but causes leakage artifacts.
Fourier transformfrequency resolutionspectral leakagewindowing
Learn - Spectrogram
The Spectrogram
A spectrogram is a rolling history of FFT snapshots: time runs left to right, frequency bottom to top, and colour encodes intensity. Dark navy means silence. Bright cyan or white means strong energy.
What to look for
Horizontal stripes are sustained tones. Vertical smears are transients such as claps or drum hits. Diagonal streaks indicate pitch glides or vibrato. Speech shows shifting formant bands separated by silence.
Adjust the history slider (1 to 8 seconds) in the sidebar. Short windows reveal rapid events; long windows show how the sound evolves over time.
time-frequencySTFTformantstransients
Learn - Harmonic Map
Harmonic Map
A vibrating string or tube produces not just f0 but a series of overtones at exact integer multiples, the harmonic series.
fn = n times f0 n = 1, 2, 3, 4 ...
Soniform plots this as a radial node graph. The central cyan node is f0. Surrounding violet nodes are overtones. Node size represents amplitude.
Two instruments playing the same note have the same node positions but very different sizes. A flute concentrates energy at n=1 and n=2. A violin spreads to n=6 and beyond. That difference is what makes them sound distinct.
overtonestimbreharmonic series