This page listens to one note at a time and describes how well it was played: how steady its pitch and volume are, whether its tone is warm or thin, and how cleanly it starts and ends. It finds each note on its own, so just play or sing a note, stop, and read the report. It's a companion to the multi-note tuner, which shows which notes are sounding rather than how they sound.

Choose an input and your instrument. For the microphone, press Enable microphone the first time to let the page use it. The microphone is processed entirely in your browser and isn't recorded or sent anywhere. To try the page without a microphone, choose the test tone, which plays repeating notes in several styles.

Microphone
Analysis
Display
Advanced settings

Pitch

About this graph

The pitch over the last few seconds, zoomed in to two semitones (or the range in the advanced settings) above and below the note you're playing, with a green line at each in-tune note: a flat line on a green line is a perfectly steady, in-tune note. The view follows your pitch as it moves, but not your vibrato. The line is dimmer where the pitch is less clear. Once a note ends, its attack is shaded amber, its steady middle green, and its release blue. A note that's still sounding is shaded gray.

On the right is the distribution of the recent pitches, in which each moment counts less the longer ago it was (how quickly is set by Distribution memory in the display settings). The green bands are ±5 cents around each note, and the line marks the most common pitch. Its width has a fixed scale: it reaches across only when the pitch has stayed within a few cents, so a pitch that wanders draws a lower, wider curve, and the curve fades away once the sound stops. Its color shows how steady the pitch is, whatever its tuning: the wobble below it is how much the pitch moves within each note (the RMS, in cents, of its movement over the last second or so, after removing slow drifts and the vibrato), and it's green up to 5 cents, amber up to 12, and red beyond. With vibrato, the distribution has two peaks, at the top and bottom of the swing, where the pitch spends the most time.

Volume

About this graph

The volume over the last few seconds, lightly smoothed and zoomed in to the loudest 36 dB (or the range in the advanced settings) so that small changes within a note are visible, with the current level at the top. The dashed line is the level of the background noise (from the calibration, if there is one), and the dotted line is the level a sound must reach to count as a note.

On the right is the distribution of the recent levels in the steady parts of notes (not their attacks and releases), weighted toward the most recent, with its spread below. Its color shows how even the volume is: green up to 1 dB, amber up to 2.5, and red beyond.

Vibrato

About this graph

Vibrato is an oscillation of the pitch, so this graph takes a second Fourier transform: of the pitch line itself, rather than of the sound. Each moment shows how strongly the pitch was oscillating at each rate (from 1 to 12 times a second) over the second around it, with brighter colors for wider swings; slides and slow drifts are removed first. A steady vibrato makes a bright horizontal band at its rate, and the dashed lines mark the rates typical of vibrato (4.5 to 7 Hz). Since each moment needs a second of pitch, the graph runs half a second behind. On the right is the recent spectrum, with its peak's rate and extent (how many cents the pitch swings above and below its center). Its color shows how even the vibrato is: green if it's even, amber if it's somewhat uneven, and red if it's very uneven. It's even when the peak accounts for most of the pitch's movement, as it does for a steady, regular swing. A rate outside the typical range is named (slow or fast) but doesn't change the color.

Attacks

About this graph

The starts of the last several notes, overlaid so you can compare them. Each is lined up at 0 ms, where its volume comes within 20 dB of the note's steady level (the dotted line). The note you're playing is drawn in amber as it happens, the report's note in white, and the other notes in blue, fainter the older they are. The top strip is the volume relative to the steady level: a clean attack rises quickly to the green line and stays there; one that shoots above it is accented or hard, and one that creeps up is slow to speak. The middle strip is the pitch relative to the note: it should land in the green band (±15 cents) right away, rather than scooping up into it or settling slowly. It's dashed where the pitch was found in a noisy sound (with a clarity below 60%), as often happens in a note's first moments. The box in the corner of each strip grades that part of the attack (its volume, pitch, or noise), with the numbers it's based on: for the note you're playing once it has lasted almost a second (with a dashed outline, since it may still change), or else for the report's note. Click a box to go to the report. A note that began by changing pitch from the previous one without a break is lined up at the change, and its boxes describe the change instead: where the pitch came from, how long it took to reach the new note (and whether it went past it), and how much the volume and clarity dipped on the way. Its range widens (up to ±200 cents) to fit a large slide, and narrows again when the recent notes stay closer to their pitch. The bottom strip is the clarity: a dip at the start means breath, bow scratch, or air before the tone.

Releases

About this graph

The ends of the last several notes, overlaid and lined up at 0 ms, where each starts to fade (its volume falls 3 dB below its steady level, the dashed line). The note you're playing appears in amber as soon as it starts to fade. The top strip shows how each note fades: a steep drop is a cutoff, and a gentle slope is a taper. The middle strip shows whether the pitch stays on the note as it fades, or sags or rises. The bottom strip shows noise as the note ends, as a dip in clarity.

Overtones

About this graph

The strength of each harmonic, in decibels below the strongest one: H1 is the fundamental (the note's pitch), H2 is twice its frequency, and so on. While a note sounds, the bars follow it live; afterward, they show the report's note. The dashed line is the overall slope of the harmonics: a steeper drop gives a darker, warmer tone, and a flatter one a brighter tone. The scale at the top places the tone between thin and warm. The gray bars on the right are the noise between the harmonics, such as breath, air, or bow noise, in three frequency bands, relative to the harmonics' total power.

Report

No notes yet

Pitch steadiness

Volume steadiness

Tone color

Noise

Attack

Release

Recent notes

NoteLengthPitchVolumeToneNoiseAttackRelease

The notes you play will be listed here. Click one to see its report.

How it works

The sound is measured every 5 milliseconds. Each step measures the volume, as the RMS level of the last 10 ms in decibels, and the pitch, using the McLeod pitch method (P. McLeod and G. Wyvill, "A Smarter Way to Find Pitch"). That method compares about 40 ms of sound with itself, delayed by \(\tau\) samples, using the normalized square difference \[ n(\tau) = \frac{2 \sum_j x_j x_{j+\tau}}{\sum_j \left(x_j^2 + x_{j+\tau}^2\right)}, \] which is 1 when the sound repeats exactly every \(\tau\) samples. The note's period is the first peak of \(n(\tau)\) that's almost as high as the highest one. The height of that peak, the clarity, shows how periodic the sound is: near 1 for a clear tone, and lower when breath, bow noise, or air is mixed in.

A note starts when the volume rises more than the detection level above the background noise. That's the median level heard while calibrating, or without a calibration, a level that the page keeps estimating between notes. It ends when the volume stays low for 80 ms, or when the pitch moves by more than 80 cents without a break, which starts the next note. Each note is then split into its attack, until it reaches full volume and a steady pitch; its sustain, the steady middle; and its release, from when it starts to fade until it's gone.

Pitch and volume steadiness are measured over the sustain. A straight line is fitted through the pitch (in cents), and its slope is the drift. Vibrato is found as the strongest regular wave between 3 and 9 Hz in what's left. It's reported on its own and isn't counted as unsteadiness. The wobble is the RMS of what's left after averaging over one vibrato cycle (or over 30 ms, without vibrato), so it measures how much the vibrato's center wanders. The volume is measured the same way, in decibels.

Tone color comes from the strength of each harmonic. Every 20 ms, a larger FFT (about 170 ms of sound, with a Gaussian window) is taken, and each harmonic's peak is located with the Gaussian interpolation described on the tuner's page. The harmonics' power is averaged over the sustain. The warmth combines two scores: how low the harmonics' center of mass is, and how strong the fundamental is compared with the strongest harmonic. This is a rule of thumb, not a standard measure; what sounds warm depends on the instrument and the listener.

Noise is measured with the same FFT. The bins in each harmonic's peak (widened when the pitch moves, as with vibrato) count as tone, and the bins between the peaks count as noise, after subtracting the background noise: the average spectrum heard while calibrating, or without a calibration, the spectrum that the page measures between notes. The noise is reported relative to the harmonics' power, in three bands: breath and air noise is mostly above 1 kHz. The power halfway between harmonics, above the noise around it, is reported as subharmonics: they appear when a vibration only repeats every second period, as with a crunching bow or a creaky or rough voice. Eventually, you'll be able to load a recording of a tone you want to imitate and compare your harmonics with its harmonics directly.

The attack is described by its rise time, from 20 dB below the sustain's level to within 3 dB of it; by how far it overshoots that level; by how far the starting pitch is from the note and how long the pitch takes to settle within 15 cents; and by how long noise is heard before the tone becomes clear. The release is described by how long the note takes to fade by 30 dB, by how far its pitch moves as it fades, and by whether its clarity drops (noise) as it ends. The thresholds that turn these numbers into descriptions are rough first guesses.