A quick definition first; open “How the app uses it” for the calculation or interpretation. You can link directly to a term, such as roughness or signal integrity.
No matching terms. Try a shorter word, or clear the search to show everything.
The sound
How many times a wave repeats each second. A 440 Hz wave repeats 440 times per second. Higher frequency usually means higher perceived pitch.
How the app uses it
For a selected note, frequency is calculated from the tuning reference and its semitone distance from A4.
f = A4 × 2^((MIDI note − 69) / 12)A4 is MIDI note 69 and C4 is note 60. At A4 = 440 Hz, C4 is about 261.63 Hz. For a file, the app estimates frequencies from peaks in its spectrum; the displayed values are estimates, not exact note transcription.
The sound
Pitch is the high-or-low quality you hear. A note name labels a pitch in a tuning system. The root note is the starting note of the chosen interval.
How the app uses it
The app maps estimated frequencies to the nearest equal-tempered note using 69 + 12 × log₂(f / A4), rounded to a whole MIDI note. A4 is a reference, not a health setting. Changing 440 to 432 shifts the tuning of selected tones; it does not prove a special biological effect.
Relationships
The distance between two pitches: unison, second, third, fifth, octave, and so on. It depends on their frequency ratio rather than their frequency difference alone.
How the app uses it
A perfect fifth is seven semitones in 12-TET, or exactly 3:2 in just tuning. The file readout summarizes the two strongest pitch classes and folds their frequency relationship into one octave. It does not identify every note in a chord.
Relationships
A comparison by division. If two frequencies are 220 and 330 Hz, the higher divided by the lower is 1.5, written 3:2.
How the app uses it
After octave-folding, the app tests reduced fractions p/q with both numbers no larger than 16. Candidates must lie within ±20 cents. It chooses the lowest value of log₂(pq) + |cents error| / 10, balancing small numbers with tuning accuracy. In Just mode, a named interval uses its chosen exact fraction directly. “None within ±20¢” means no candidate met this rule.
Relationships
A precise unit for pitch distance. One equal-tempered semitone is 100 cents; one octave is 1,200 cents. The ¢ symbol here means cents of pitch.
How the app uses it
distance = 1200 × log₂(f₂ / f₁)
error = 1200 × log₂(actual ratio / target ratio)A 3:2 fifth is about 701.96 cents; a 12-TET fifth is 700 cents. Positive error means sharper than the target ratio, negative means flatter.
Relationships
12-tone equal temperament divides each octave into 12 equal pitch steps. Just tuning builds the selected interval from a whole-number frequency ratio.
How the app uses it
12-TET: f₂ = f₁ × 2^(semitones / 12)
Just ratio: f₂ = f₁ × p/qThe tradeoff is musical: equal temperament makes the same interval structure available in every key; just ratios give exact relationships for the chosen pair. Neither makes every kind of music universally better.
Relationships
Nearby frequencies can create a pulsing loudness envelope. The app’s readout specifically estimates the mismatch between harmonics associated with the selected simple ratio.
How the app uses it
app beat rate = |q × f₂ − p × f₁|For a 3:2 ratio, compare 2 × the higher note with 3 × the lower note. At 220 and 331 Hz, that mismatch is 2 Hz. Ordinary beating between two nearby fundamentals uses |f₂ − f₁|; these are different comparisons. With pure sine tones, the higher harmonics are absent, so this remains a calculated relationship rather than a measured audible pulse. File analysis suppresses mismatches below estimated frequency-resolution error. This is not BPM, and the app does not calculate binaural effects.
The sound
Partials are the frequency components inside a sound. Harmonics are partials at whole-number multiples of a fundamental: f, 2f, 3f, and so on.
How the app uses it
“Sine” gives each generated tone one component. “6 harmonics” gives it six, with amplitude 1/n for harmonic number n. Coincident components are merged. In audio files, strong peaks may be harmonics, separate notes, or other sound components; a detected peak is not automatically a fundamental.
The sound
The character of a sound: why a flute and a violin can play the same note and still sound different. Their partials, loudness envelopes, and noise differ.
How the app uses it
The tone controls compare sine waves with a six-harmonic model. Adding harmonics creates many more interactions between nearby components, changing roughness and the dissonance curve. A simple frequency ratio alone cannot predict the sound of every instrument.
The reading
A model of the sensory friction created by nearby frequency components. Slow interference can sound like pulses; closer, faster interactions can sound rough.
How the app uses it
For each pair of frequency-sorted peaks, the app applies a Sethares-style curve, weighted by the smaller amplitude, then divides the sum by total amplitude.
s = 0.24 / (0.0207 × f_low + 18.96)
x = s × (f_high − f_low)
pair = min(a_low, a_high) × (e^(−3.5x) − e^(−5.75x))The displayed roughness is divided by a reference: C4 plus an equal-tempered minor second, each with six harmonics. 1.00 means that reference; it is not a percentage or a universal limit. Absolute pitch and timbre affect the result. Read Sethares’ explanation of timbre and consonance.
The reading
A 0–100 estimate combining low roughness, a harmonic spectrum, and relatively simple pitch ratios. It summarizes this model’s idea of consonance.
How the app uses it
The weights are 50% low roughness, 30% harmonic energy, and 20% low ratio complexity. Scores of 70 or more are “Harmonious”; 40–under 70 is “Balanced tension”; below 40 is “Tense / dissonant”. The full formula is below. A score does not measure musical quality, emotional value, spiritual significance, or health.
The reading
A graph of modeled roughness as the second tone moves from unison to an octave above the first. A dip marks a relationship with lower roughness for that timbre and register.
How the app uses it
The app evaluates two tones with the selected timbre across 0–1,200 cents, using 5-cent samples plus exact simple-ratio positions. With harmonic tones, shared partials can create dips. With pure sine tones, named intervals do not create the same harmonic coincidences. The moving dot marks the current interval; it is not a recording of a listener’s preference.
Audio cues
A separate 0–100 estimate built from signal level, spectral brightness, and spectral change. “May evoke” labels offer possible associations with those cues.
How the app uses it
Energy uses 70% normalized RMS level, 20% normalized brightness, and 10% positive spectral flux. Below 35 is “Low vibe”; 65 or more is “High vibe”; the middle is “Balanced vibe”. Bass share is supporting evidence and does not directly add to this score. Roughness qualifies the suggested feeling labels, but does not enter the energy score. These weights are an engineering heuristic, not a validated emotion detector or a measure of spiritual energy.
Audio cues
The power-weighted average frequency of a spectrum. More power at high frequencies raises this value and often corresponds to a brighter sound.
How the app uses it
centroid = Σ(frequency × power) / ΣpowerThe app excludes the zero-frequency bin and normalizes the centroid on a logarithmic scale from 150 to 6,000 Hz for its energy estimate. Brightness is not the same as the strongest note’s pitch, and it does not say whether a sound is pleasant.
Audio cues
The percentage of analyzed spectral power at frequencies up to 250 Hz. It describes the low-frequency share, rather than the overall volume.
How the app uses it
bass % = 100 × power at ≤250 Hz / total spectral powerThe app derives power from the spectrum, excludes the zero-frequency bin, and displays the share as supporting audio evidence. A high bass percentage can belong to a quiet signal. This number does not directly contribute to the energy score.
Audio cues
How much the distribution of frequency power has shifted toward newly stronger components since the previous spectrum. It can rise around changing notes or transients.
How the app uses it
Each spectrum is normalized to total power 1. The app sums positive bin-by-bin changes: Σmax(0, current share − previous share). The result lies between 0 and 1 and is displayed as a percentage. Cue features update at up to 15 times per second. This is not tempo detection; stationary analytic tone previews use zero flux.
The reading
The fraction of detected peak energy that fits up to two distinct harmonic families. A family contains approximately whole-number multiples of a possible fundamental.
How the app uses it
The app proposes fundamentals from the five strongest peaks and their divisions by 1 through 4, keeping candidates at least 40 Hz. It selects up to two families, adds only uncovered peaks, and accepts harmonics within ±25 cents. The final fraction is covered Σa² / total Σa². Generated interval tones use a value of 1. Dense chords with more than two families can score lower even when they sound pleasant.
The reading
A way to describe the size of a reduced frequency ratio’s numbers. Smaller numbers usually mean lower ratio complexity in this model.
How the app uses it
Tenney height = log₂(p × q)A 3:2 ratio has height about 2.585; a 16:15 ratio has height about 7.907. For a file, the app averages pairwise heights of the four strongest peaks, weighted by their amplitude products, after octave-folding. If a pair has no nearby ratio, it uses the reference maximum of log₂(240), about 7.907. This is a mathematical measure, not a judgment of a listener.
The measurement
A spectrum shows how much of each frequency is present. An FFT is a calculation that turns a short slice of audio into frequency bins. Peaks are prominent local maxima.
How the app uses it
The app uses 8,192-sample, Blackman-windowed frames. Bin spacing is sample rate / 8192: about 5.38 Hz at 44.1 kHz. Interpolation refines peak locations but does not remove uncertainty. It selects up to 24 peaks from roughly 40–9,000 Hz, above an estimated noise floor +6 dB and within 60 dB of the strongest peak. The search also rejects very weak spectra. Longer windows trade better frequency detail for coarser time detail.
The measurement
The waveform shows the recording’s amplitude over time. A spectrogram shows its frequency content over time: horizontal position is time, vertical position is frequency, and color represents strength.
How the app uses it
The track scan produces a min/max waveform outline, a log-frequency spectrogram, and one-second harmony summaries. Bright lines may be notes or harmonics; dense bands may be complex tones or noise. The harmony line is a separate model result, not the spectrogram’s brightness.
The recording
A check for certain problems in the recorded signal, including flat-topped peaks, little headroom, DC offset, and strongly opposing stereo channels.
How the app uses it
High-severity clipping or stereo phase flags produce “Distorted”; lesser flags produce “Minor issues”; no detected flags produces “Clean”. Integrity is calculated separately from harmony and does not directly change the harmony formula. “Clean” only means these checks found no issue; it does not certify a recording or detect every kind of distortion.
The recording
When a signal exceeds a system’s available range, its peaks can flatten. That changes the waveform and can add unwanted frequency components.
How the app uses it
The detector looks for runs of at least three samples at the decoded file’s absolute peak, within a tolerance of 0.000001, when that peak exceeds 0.05 of full scale. It raises a high flag if flat runs cover more than 0.01% of all channel samples or any run reaches eight samples. This is a heuristic: lossy encoding can change flat tops, so MP3/AAC clipping cannot be checked exactly with this method.
The recording
Headroom is the margin below a digital limit. dBFS expresses a signal level relative to full scale. RMS describes its average magnitude rather than its single highest sample.
How the app uses it
sample peak dBFS = 20 × log₁₀(max |sample|)
RMS dBFS = 10 × log₁₀(mean(sample²))A sample peak above −0.1 dBFS gets a low headroom flag. Track frames below −50 dBFS RMS are omitted from harmony scoring. These are digital levels; the app does not measure sound pressure at your ears or the room’s volume.
The recording
A waveform’s baseline sits above or below zero instead of being centered. It uses some of the available signal range.
How the app uses it
The app calculates each channel’s mean sample value. If the largest absolute mean exceeds 0.005, or 0.5% of full scale, it reports a minor DC-offset issue.
The recording
How similarly the left and right channels move. Strongly opposite channels can cancel when combined into mono, changing what the analyzer sees.
How the app uses it
The app compares the first two channels using normalized sample products, sampling every fourth sample. A value near +1 means similar movement; near −1 means opposing movement. A result below −0.3 raises a high phase flag. Spectral analysis uses the mean of all channels; wide stereo can therefore produce different results from each channel on its own.
The picture
Cymatics explores visible patterns associated with vibration. Here, the picture is generated by a simulation or mathematical drawing driven by selected or estimated sound relationships.
How the app uses it
Resonance Garden draws a designed circular wave field. Sand models particles settling toward low-energy regions of a driven square plate. Water uses a wave model. Harmonograph, phyllotaxis, rose, and Turing modes use different drawing or growth rules. Spectrum Bloom and Spectral Terrain plot frequency power; Wave Interference and Fourier Orbits build illustrative waves from partial magnitudes. Compare the views and their assumptions. Shape depends on these rules and model assumptions, as well as sound. It is not an image measured from your speaker, body, or a physical water experiment.
The picture
A description of how a picture matches itself when rotated or reflected. C4 suggests fourfold rotational symmetry; D4 adds reflection symmetry. C1 means no higher rotation was accepted.
How the app uses it
The app samples a 64 × 64 pattern field, tests rotation orders 2, 3, 4, 5, 6, 8, and 12, and checks mirror matches. A correlation of at least 0.78 is required. “Corr” measures resemblance between compared pattern samples; it is not a confidence probability or a measure of musical harmony. Untested orders can get a lower-order label. The square plate model can favor fourfold geometry; symmetry is not evidence of a hidden message.
The picture
A pitch class groups notes with the same name across octaves, such as every C. The pitch clock places the 12 classes around the circle of fifths.
How the app uses it
Detected peak energy is grouped by the nearest equal-tempered pitch class. The readout keeps up to four classes, each at least 15% as strong as the strongest. File playback smooths these weights over time. The pitch clock highlights them; Metatron overlay uses pitch-class weights to light selected connections. Other overlays use chosen geometric constructions or detected pattern symmetry. Resemblance to a symbolic diagram is not an independent discovery in the audio.
The picture
The golden ratio is φ = (1 + √5) / 2, about 1.618. Phyllotaxis refers to arrangements of leaves or seeds; a golden-angle spiral is one mathematical model of such arrangements.
How the app uses it
The φ interval sets the second tone to φ times the first; φ is irrational, so any fraction label, such as 8:5, is an approximation. Phyllotaxis mode places points using a turn angle derived from the fractional part of the frequency ratio. With φ, the app turns about 222.49° per point, equivalent to a reverse turn of the familiar 137.51° golden angle. Repeating a useful mathematical rule produces a pattern; it does not establish a special healing frequency or a universal law for every living structure.
The picture
Physical resonance is a strong response when a system is driven near a natural mode. Resonance Garden is the app’s designed circular wave visualization.
How the app uses it
The garden uses up to six strong, spatially resolvable peaks in a 4 m circular model, with an assumed wave speed of 343 m/s. Frequencies set wavelengths; small ratios help choose angular petal structure; amplitudes set wave weights. Bessel functions describe the radial shapes. Phase and a harmony-dependent coherence blend are artistic mappings, and wave time is slowed 512 times for display. These choices explain the geometry; the picture does not measure an actual resonator or cosmic energy.