The Laryngotron is a monophonic voice synthesiser with none of the controls a synthesiser normally has. In their place sit thirteen levers, each one a structure of the larynx and vocal tract, and each wired to the audio rather than to a label announcing that it is.

A sawtooth glottal source is shaped by the body–cover and cartilage figures, passed through five formant resonators that the larynx, tongue, jaw and lips move about, and mixed with a parallel nasal branch that opens when you drop the velum. Constricting the false vocal folds engages a waveshaper, which is approximately what constricting them does to you.

Every option is a discrete state rather than a continuous knob, exactly as in the training. You are not blending between settings. You are choosing one and then living with the consequences, which is also the situation on stage.

True vocal fold onset, false vocal folds, true vocal fold body–cover, thyroid cartilage, cricoid cartilage, larynx height, velum, tongue body, aryepiglottic sphincter, jaw, lips, head and neck anchoring, torso anchoring. Thirteen things you have been operating your entire life without being introduced to any of them.

Onset offers glottal, smooth and aspirate. Body–cover offers slack, thick, thin and stiff — four adjectives nobody has ever volunteered about their own voice. The cartilages are vertical or tilted. Larynx, velum, tongue, sphincter, jaw and lips each have three positions. The anchors are on or off, and are the only two that make no sound at all.

Everything is reflected at once in the audio, in an animated cross-section of a head, and in a numeric readout giving source brightness, aspiration percentage, the first two formant frequencies, the aryepiglottic peak and its gain, the nasal branch level, and your anchoring set against how much effort you are currently demanding.

Body–cover is the source itself: the lowpass on the glottal waveform, the aspiration floor and the level. Slack is dark and leaky. Thick is rich and loud. Thin is lighter. Stiff is very nearly a sine wave with an air leak, and is what most people do when asked to sing quietly.

Tilting the thyroid cartilage thins the cover — brightness falls to 78% and level to 92%. Sweeter, easier at the top, and quieter, which is precisely why nobody has ever belted a lullaby.

Tilting the cricoid cartilage does the opposite and does it enthusiastically: brightness up 45%, a 5 dB shelf above 2.2 kHz, level up a tenth. It is the only figure unique to a single recipe, and that recipe is Belt, and it would like a word with your torso.

Onset controls the amplitude envelope — four milliseconds with a small overshoot for glottal, fifty for smooth, a hundred and thirty for aspirate, the last with extra breath. One of these is how people begin sentences when annoyed.

Larynx height scales the entire tube. Every formant is multiplied by 0.86, 1.0 or 1.16, because changing the length of a resonator moves all of its resonances at once — which is also why the impression of a small child is done with the larynx and not the throat.

Tongue body moves the first two formants in opposite directions: a high tongue takes F1 down by a fifth and pushes F2 up by more than a third. Jaw and lips make smaller adjustments, the lips by lengthening or shortening the open end of the tube, which is the entire physics of pouting.

The velum opens a real second acoustic path at nought, twenty-two or fifty-five per cent: a resonance at 270 Hz with an anti-resonance above it. It is the only lever that adds another route through the instrument rather than colouring the existing one, and the only one that can install a small Frenchman in your nose.

The aryepiglottic sphincter is a peaking filter near 3 kHz — six decibels down when wide, flat at mid, fifteen decibels up when narrow, plus a further lift in level. That happens to be the frequency human hearing is most sensitive to, so narrowing it makes you dramatically louder for no additional work. It is the loudest cheap trick in singing and it is free.

The two anchoring figures change no timbre whatsoever. What they change is what the instrument will let you get away with.

A running tally of effort demand is kept — one unit each for thick body–cover, a narrow sphincter, a high larynx and a tilted cricoid. One unit is free, because that is merely talking. Each anchor engaged offsets 1.6 units. Whatever is left over becomes audible wobble.

So an unanchored belt shakes and an anchored one does not, while the two are timbrally identical. There is a gauge, and it reads effort in the wrong place. It responds to constriction and to unsupported demand, and it ignores loudness entirely, which is the opposite of what most people assume it should do.

Constrict the false vocal folds and the sound gets harsher and quieter simultaneously — more work, less voice — at which point a Health & Safety Inspector slides in from the corner citing you a Form 7B. He is a joke. What he is complaining about is not: constriction is interference rather than effort, and it is the one thing on the panel that is genuinely bad for you. A Silent Laugh button dismisses both of them.

Six voice qualities, with twang given in both its oral and nasal forms, because they differ by exactly one lever and that is far too good a demonstration to waste.

• Speech — glottal onset, thick folds, both cartilages vertical, wide sphincter, nothing anchored. The one you already own.
• Falsetto — Speech with two levers moved. Structurally honest, dynamically hopeless, wildly popular in stairwells.
• Sob — thyroid tilted, larynx low, thin folds, both anchors on. Instant Sunday-night drama documentary.
• Oral Twang — larynx up, sphincter narrowed, no anchoring required at all. Twenty decibels, no invoice.
• Nasal Twang — Oral Twang with the velum at mid. One lever, entirely different genre.
• Opera — Sob plus a narrowed sphincter, fully anchored, cricoid still vertical. Congratulations, you sound expensive.
• Belt — thick, high, narrow, tilted, anchored at both ends. Thrilling. Not to be attempted while distracted.

A tour control plays all seven in sequence on one unchanging pitch, captioned as it goes. A lever marked Pull for a new quality randomises all thirteen figures, plays the result and names it, which is how the world was introduced to the Altitudinous Adenoidal Full-Fat Piercing Whinge.

A live mid-sagittal diagram animates with every lever. The larynx rides up and down, the cartilages rotate, the aryepiglottic aperture visibly closes, the velum swings across the nasal port, the vocal folds thicken and thin, and green struts appear at the neck and torso when you anchor. The false folds turn red when you misbehave.

A spectrum analyser sits below it. Narrow the sphincter and a hump appears near 3 kHz and stays there while you play up and down, which is the shortest available explanation of the difference between a formant and a harmonic.

Notes are monophonic and legato, because you have the one larynx. The keyboard is at the bottom of the interface, deliberately undersized, labelled the least interesting parameter. A siren control glides an octave and a half on the current setting, in case you would like to find out where your recipe stops working.

A full manual is built in and opens with the ? key. Its figure table and recipe grid are generated at open time from the same constants the synthesiser reads, so the documentation is incapable of describing an instrument that no longer exists.

Underneath the brass fittings and the Inspectorate there is a serious question, which is whether the relationship between vocal anatomy and vocal quality is easier to hold onto when it can be operated rather than described.

Voice teaching runs on metaphor — support, placement, mask, spin — because the structures involved are invisible, unfamiliar and awkward to feel. Metaphor works until two people discover they have been meaning different things by the same word for a year. An instrument where each structure is separately audible is a different sort of argument, and a faster one.

The satire is aimed at the mystique rather than at the method. Singing is routinely discussed as a gift, a calling, or a thing you either have or do not. The joke of this instrument is that the sublime turns out to have thirteen settings, several of which are called stiff. That is funny, and it is also true, and the second part is the reason it is worth building.

The claim being tested is a strong one: that quality is independent of pitch, and that effort in the right place is a different thing from effort. Both are demonstrable here in about four seconds each, which is a considerable saving on a term.

Stated plainly, because this will be shown to people who know more about voices than the machine does.

The source is a filtered sawtooth, not a glottal flow model — body–cover changes spectral tilt convincingly but is not simulating vocal fold collision. There is one vowel, whose formants the tongue, jaw and lip figures shift; it is not a full vowel space. The formants are peaking filters in series and do not model the tract as a tube. Onset options alter the amplitude envelope only. Register transitions are not modelled at all, so nothing here will break at a passaggio, which is the single respect in which it is easier than a person. There is no microphone input: it cannot hear you and does not pretend to diagnose you.

The recipes follow the published Estill figure combinations. Wordings and individual options vary between editions and between teachers — the sphincter setting for Speech is given as wide in the source used here and as mid elsewhere, and Opera’s body–cover is conventionally thick for men and thin for women, where this instrument uses thick. If a recipe here disagrees with the room you are standing in, the room is right.

An independent teaching toy. Not affiliated with, endorsed by, or produced in association with Estill Voice International. Not a substitute for a Certified Master Teacher, a mirror, or a nasendoscope.