All the strokes

The vocabulary is style-independent: the DNA picks a name, the style decides how it draws it. On this page every name appears once – drawn by exactly the code that also draws the sheet.

The distribution is not uniform. It sits in the catalogue, and it is deliberately skewed: the variety lies in the eyes and mouths, because that is where a viewer reads character. Hairdos and headwear stay small – a sheet full of hats is a hat shop, not a sheet full of people.

Eyes – sixteen

Every shape reaches into the same local ellipse: the gaze shifts it, the lid squashes it. That is why the lid factor stands in every eye shape. The pupil size comes from the DNA – a needle and a button in the same eye shape are two different people.

monolid is a lid without a crease. The upper edge is one line: it rises gently from the outer corner and drops more steeply towards the inner one, where a short epicanthic fold covers the corner. That is what separates it from mandel, where two even arcs meet in two sharp corners. The iris sits high and is cut by the lid instead of floating in the opening. One of the most common lid shapes in the world – its absence here was a gap in the vocabulary, not a decision; with weight 1.5 it is the second most common shape after offen.

leer, empty, is the ring without a pupil but with a second line around the outside of the eye – a contour drawn twice. Nothing is filled, and that is exactly the statement: a gaze with nothing in it. The outer ring is thinner and paler than the inner one; two lines of equal weight read as a target, this way it reads as a line somebody went over a second time. And it does not travel with the gaze – it belongs to the eye, not to the gaze, and there is nothing here that could look.

Nor does it sit exactly concentric, but off by a hair – and differently for each eye, in direction and amount. The first state used the same offset for both, arguing that a second printing plate slips as a whole. That holds for a plate and not for this drawing: here a hand goes over the line a second time, and a hand misses twice in two different ways. The same offset on both eyes reads as intent, different ones read as drawing.

It is rolled from a side stream of its own (handStrom(dna, 'auge', 'leerversatz')), not from the handwriting stream. A kind that arrives late must not shift the handwriting of every other head – a side stream hangs on a name, not on a position.

The only shape without an eye shape is funkeln, sparkle. No ring, no lid, no pupil – two big stars, each with its own line around the fill. Without that line a coloured shape would stand alone in a face that is drawn everywhere else; fill and line have their own tracks and therefore wobble differently, as in a coloured-in drawing.

Everything about it scatters per side: number of points (four or five, five a little more often – two four-point stars in one face never happen, that would be two crosses), point lengths, angular offset and size. The size runs from barely an eye width to more than two – capped by the room up to the silhouette, because a point sticking out over the edge of the head is no longer a big star but a mistake. When the head blinks no lid closes (there is none): the star squashes instead. funkeln has weight 0.16, the rarest eye shape – two big coloured stars are the loudest head on the sheet.

It is also the only eye shape with a layer of its own. Every other one sits in the face and is covered by the fringe like the forehead beneath it; the pure star is not an eye shape but a mark on the head – under a hairdo it would sit there clipped. So it lies above hair and headgear, and only the glasses lie above it. The brow does not travel with it: it is a line on the skin and belongs under the fringe. The star then covers it – whatever is big enough to sit above the hairdo is big enough for the brow as well. See The drawing order.

The highlight turns a hole into an eye. It is left out, not punched: elongated, tilted across the light, in skin colour – and without a seam. A pale rim on the dark pupil would look like a stain, not like light.

The offset of the highlight lies in the feld too: otherwise it slips out of the squashed pupil when the head is turned. The same field tilts the pupil itself along with the surface it lies on (see The pen).

The pupil may travel within the white, but not over the edge: its path is (rr − pr), the radius minus its own size. With the lid closed it stands still – otherwise it visibly slides out of the eye during a blink.

Above lider > 0.72 an eye is always the same arc, no matter which shape. A blink does not have thirteen variants. zwinker is the only feature in the whole face where the two sides differ – which is why it stands in the shape function and not in a symmetry rule further up.

The star out of the character set

Symptom: The star eye looked like a symbol somebody had pasted into the drawing.

Cause: Four points of equal length at right angles. The compass puts them exactly where one expects them.

Fix: wuerfleStern per head: four or five points, each scaled by 0.8 to 1.12, each offset by up to 0.14 in angle, and the notch between them sits in the middle of its two neighbours – not on the fixed half step, or it would drift towards a point when they are offset.

Addendum: The star had a fixed size of 0.4 eye radii and stayed neatly inside the ring at 0.7 – a button with points. Now wuerfleSternGroesse scatters between 0.44 and 0.72, with outliers down to 0.36 and up to 0.92. The big ones reach over the ring with their points; the fill sits on top of the stroke, not beneath it. The star may only wander as far as it is smaller than the ring.

Mouths – twelve

A mouth has two states: the one from its DNA and the open one. The open one belongs to no shape – a talking mouth looks the same on everybody. That is why it stands before the case distinction, and above oeffnung > 0.16 all twelve shapes are an ellipse with a tongue.

The four closed base shapes (strich, laecheln, klein, welle) are not drawing steps but a data table LINIEN. There is a single reason for that: the heart needs one of them as a backing and has to know where its apex lies.

How far a mouth runs

The width has two dials: the scale of the face – it has always had that – and, since 21.08.2026, a mouth width per head (0.8 … 1.35, from the handwriting stream). Before, it hung on the scale alone, which meant every mouth was equally wide in relation to its head: a sheet full of mouths in the same proportion.

Outwards the mouth ends at the cheek zone, and it may now run right up to it. The old cap let it fill only half the zone width – a grin was a little stroke in a big face.

The heart is an overlay

lippen is not a mouth shape of its own. Underneath lies a perfectly ordinary closed line, and it runs past the heart to the left and right – the way one draws lips without reinventing the mouth. The wave is ruled out as a backing: two curves on top of each other are two statements.

The pancake under the line

Symptom: On some faces the heart was no longer a heart shape but a broad blot under the mouth.

Cause: The height of the heart was measured against the mouth height. On a flat face that is a fraction of the mouth width – the heart grew wide and was squeezed flat.

Fix: Both axes measure against the width. Plus three rolled measurements per head: width, height, depth of the notch. Incidentally, a deep notch turns the heart into a tulip – which is why it only goes up to 0.85.

Three waves instead of two arcs

Symptom: With a small notch the top edge of the heart was not round but a row of three humps.

Cause: Peaks (±0.26) and flanks (±0.1) sat as fixed points in the list, and only the midpoint hung off the notch. Below about 0.4 the middle therefore stood higher than its own flanks: beside the two arcs a third one grew.

Fix: The top edge is one arc into which the notch cuts as a bell (herzPunkte). The bell runs out at its edge with zero slope, so it sits on the arc without a kink; at kerbe = 0 a round dome remains, and the shoulders of the arcs only come about through the incision.

And the heart sits centred on the apex of its line, not on the mouth anchor: otherwise it would hang in the air under a smile.

Teeth need a surface

zaehne is a lens – almost straight on top, the arc below – filled in skin tone, with four tooth lines running top to bottom. hasenzahn is two rectangles sitting on the lip line and jutting downwards, also a surface in skin tone with an ink border.

Without a surface the grid reads as a second mouth and the buck tooth as a fence. The three tooth lines that once stood in grinsen were, by contrast, too short and read as lint – teeth now have a shape of their own.

The laughing mouth is a kidney

lachen is the wide open mouth: the top edge rises to a shoulder between the middle and the corner and sags again in the middle – the notch of the kidney – while the bottom edge bulges downwards. It measures itself in mouth widths, not in mouth heights: on some faces that is a sixth of the width, and measured against it the wide open mouth would be a slit.

It shows tongue or teeth, never both – one roll, three cases, the third being the bare hole. The tongue is a heart point down, and it is the same heart this head would wear on a lippen mouth: a head has one handwriting, not two. It does not grow along – it is a piece of body, not a share of an opening, which is why a wide open mouth has a lot of black around a small tongue.

The teeth are a band hanging from the lip, running from corner to corner. On top it follows the lip including its notch – straight teeth under a sagging upper lip would be a foreign body – underneath only halfway: fully would be a second lip, and the row of teeth would become a crescent moon. It runs out at the corners, and its lower edge is clamped against the lower edge of the mouth: the tooth stops where the mouth stops.

The roll only applies in the middle. Below 0.42 mouth widths the laughing mouth shows nothing – no tongue fits into a mouth that size, it fills it completely – above 0.9 always one of the two: a wide open mouth with nothing in it is a hole in the paper. Which of the two is decided by the same number that otherwise makes the tongue wider or narrower; there is already a roll at that spot, and a second would be one too many.

The mouth that ran off the face

Symptom: On every second head the laughing mouth stood outside the silhouette – past the cheek at the sides, past the chin below.

Cause: Two mistakes on top of each other. Every feature is drawn in the tangent plane of its anchor; that is right as long as it is small. The laughing mouth is not: drawn flat it keeps running straight while the surface below it tips away. And its limit was the mouth zone – space on the sphere, which knows nothing about the silhouette in the image.

Fix: It lies on the sphere (kugelFlaeche: every point gets its own field), and its width is measured against the outline (lachRaum). Measured across 400 heads in four poses: before, 608 of 1600 images crossed the contour, by up to 38 %; afterwards five, by one percent.

The breathing mouth

Symptom: The mouth fitted head-on but stuck out of the turned head – and measuring it per image, it grew and shrank while the head followed the pointer.

Cause: The size of a feature must not depend on the pose. Measured against the outline it depends on exactly that, because the outline turns along.

Fix: Once per head, at the corners of the pose space – three yaws × three pitches × two rolls – and the tightest one wins. If the room down to the chin is not enough, the kidney becomes flatter rather than narrower: the width is the feature.

The astonishment that turned into a yawn

Symptom: After the mouth zone had grown, the old offen was no longer a round O but a wide ellipse across the face.

Cause: The ellipse computes in mouth widths (rx = 0.3). A hole of 0.3 mouth widths grows with every mouth width – as long as the mouth width was tightly capped, nobody noticed.

Fix: A cap of its own for this mouth (OFFEN_MAX = 0.3), set at the mouth width and not at the ellipse. It cannot go lower: the mouth width has its floor at 0.24, and a cap right above it would give every head the same circle.

The beard that grew along

Symptom: Once the mouth was allowed to run further, the chin stubble was suddenly much larger.

Cause: The beard was handed the mouth width as its measure – sensible as long as the mouth width hung on the scale of the face alone. With the per-head mouth width and the laughing mouth it suddenly hung on something that has nothing to do with the beard.

Fix: A beard measure of its own (gesichtsMasse().bart): the mouth as it stands at rest in the face. A wide mouth makes a wider smile, not a thicker beard.

Noses, brows, cheeks, marks, beards, collars, trimmings

Noses

A nose is an open stroke, not an outline – except for knopf, and that is a closed arc, not a circle. It points in the direction of the gaze: a head turned to the right does not have a nose pointing left.

On top of that it leans with the rotation – lehn = yaw * 0.5, applied to the local x coordinate as a function of height. The nose stands out from the face; it therefore tips further in the picture than the skin it sits on.

Brows

A raised brow is the cheapest and clearest sign that somebody has noticed something: wach lifts it by up to 0.38 brow heights. Together with the stronger cheeks and the head that rises a little, that is the entire reaction to the pointer – more would be fidgeting.

The direction tells the rest: schraeg is low inside, high outside – the appraising look. sorge is exactly the other way round.

Cheeks

rosig is five scribbled lines of different lengths, offset, set by hand – no row, no rings. They are the trademark of the style and become stronger with attention, as if somebody were blushing.

punkte is the classic red dot: round, but by hand – slightly oval, a little tilted, with the pen's wobble. Its size is rolled per head, from a dab to an apple.

Freckles have a colour of their own (sprosse, a warm tone between red and brown), not ink: grey dots on the cheek read as stubble or dirt.

The clock face on the cheek

Symptom: Seven freckles looked like a clock.

Cause: Seven dots on a ring at equal angular spacing are a clock face, and the eye recognises it immediately.

Fix: The golden angle (2.399 rad) with a growing radius – the same arrangement as seeds in a sunflower, which never forms a pattern. On top of that every freckle gets its own trace: seven identical shapes would be seven stamps.

The cheeks sit on the cheek, not at the ear (min(0.9, augeU + 0.22)): further out the surface turns away, and five lines become one. And they keep their distance from the eye, even a large one.

Marks

Eye rings, hatching at the temple, forehead lines. All pale, thin – and ruhig, calm. Fine secondary lines that boil along do not read as a tremor of the hand but as a flicker of the picture.

The eye ring has one arc, and in a good quarter of cases two (ringBoegen in the head data). Two arcs on every head were a pattern, not a feature. The arc lies at the eye, not on the cheek: its ends stand at the height of the lid corners, in the middle it dips just below the eyeball, and it is wider than the pupil – a narrow arc far below is a laugh line.

The hatching sits on one temple. On both sides it would be a pattern, on one side it is a quirk. Which side is stored in the style's head data and not in the DNA – it is handwriting, not identity.

The fourth mark is not a secondary line: the cheek arc. The cheek pushes into the eye from below and covers its lower edge – a squint or a grin. It is the counterpart of the eye ring (cheek up instead of skin hanging) and therefore stands in the same category: a face has one or the other, never both. And it needs an eyeball it can cut into; with a line, a cross or a laughing arc it drops out of the table in the catalogue (WANGENBOGEN_AUGEN).

It is not drawn on the skin but inside the eye stroke: the eye is drawn through a window above the arc (ctx.clip), the arc on top. A skin surface used as an occluder would have to reproduce the shadow of the head surface and would lie over the cheeks; a clip paints nothing, it leaves out. The line covers the cut edge because it is wider than its own wobble, and the clip follows the same abtasten curve as the line. During a blink it fades out and is missing entirely as soon as the eye is closed – a second arc under the closed lid would be a double lid.

Four scatters per head, from the head data: how high the dome juts into the eye (from round to almost straight), how far it runs out at the side, how strongly it tilts (high outside, low inside – the cheek comes from below and outside), and whether it sits on both cheeks or only on the mark side. With the open eye and its pale lower lid the arc is thinner and paler (AUGAPFEL[typ].sanft): the cheek must not be stronger than what it covers.

The stump and the hooks

Symptom: On narrow heads the arc ran past the contour. Cut off, a stump remained that looked like a line under the eye. And the arc itself read as a straight with two hooks at its ends.

Cause: No fixed run-out measure fits every skull – on narrow ones the eyes sit close to the edge, on wide ones far from it. And a half ellipse drawn all the way to π and 2π falls steeply at the flanks; the dome is flat, the ends kink.

Fix: The arc is checked against the outline drawn in by six percent; if it does not fit it becomes narrower as a whole, right up to the eyeball – and is missing if even that is not enough. Only the cap of the ellipse is drawn (1.15π to 1.92π), stretched by 1/cos(0.15π): continuous curvature, one oval.

Beards

The moustache is a MASS, not a line: two wings in hair colour with an ink border, wider than the nose. Each wing arches up from the middle, falls away outwards and runs out into a point; a notch remains in the middle.

The zwirbel has the same wings, but each runs out into a tail that turns outward and then upward into a small curl – nothing nestles there, or the curl would be no curl. The walross is not a pair of wings but a mass hanging from the nose: narrow under the nose tip, widening downward until, at 1.75 mouth half-widths, it reaches well beyond the mouth; at the bottom it follows the top edge of the mouth halfway, stays clear above the corners, drops slightly beyond them and leaves a strip of skin to the line, inside a few strands. It does not hang over the mouth line – that would take a layer above the mouth – and it does not exist over the heart: a dark mass close above a red patch reads as a gaping mouth with a tongue.

The stoppeln span from the mouth to the chin: each dot starts three guards below the bottom edge of the mouth at its own position (beside the mouth at its lowest point, or they would climb to corner height beside a smile) and the field reaches the lower pole of the skull, narrowing downward. Every third dot grows stronger toward the chin. Before, the field lay at a fixed depth under the anchor – on a long face it hung in the middle of the cheek.

The moustache guesses the mouth

Symptom: The moustache sat right over a straight line, too low over a smile, inside the open mouth – and the mouth line painted over it, because its layer is above.

Cause: The beard knew one number about the mouth: "0.18 mouth heights above the anchor". The mouth itself was a switch of strokes, not a shape one could read from.

Fix: mundForm – the mouth as data, see above. The beard reads the top edge of this mouth, the stubble the bottom edge. The same pattern as NASEN_TIEFE, only without a table: the shape exists before anything is drawn.

The walrus grows out past the chin

Symptom: On a head with little room between nose and mouth the walrus ran as a bar down below the neck.

Cause: Its droop outside the mouth corners was divided by the distance "beard width minus corner distance". With a small b the corners lie further out than the beard reaches, the distance turns negative, and the fraction becomes a large number with the wrong sign.

Fix: The bottom edge is tied to the mouth width, not to b, and fractions are clamped. A fraction whose denominator depends on two quantities needs a floor.

The difficult question is not its shape but its place.

The beard in the nostril

Symptom: With long noses the moustache grew into the nose, with low-sitting mouths over the upper lip.

Cause: The computation used the layout zones. But the nose zone reaches all the way to the mouth zone – it says where a nose may be, not where this nose actually ends.

Fix: NASEN_TIEFE – how far each nose shape juts below its anchor, read off the drawn strokes in zeichneNase. Between them a protected space of 0.03 that no beard enters. When the space gets tight, the beard shrinks.

The price of this solution stands as a comment above the table: if a nose shape changes, the value has to change with it. A drawing that measures itself has two places that must match – the alternative would be collecting the drawn points at runtime, and that costs more per frame than the nose itself.

For the mouth that price has since been paid, in a different way: the mouth is a shape first and a drawing second. mundForm computes, per head and per frame, its top edge, bottom edge, outline and corners; zeichneMund draws from it, and the beard reads from it. The lower mark is the top edge of this mouth above its middle plus the protected space – no longer a fixed number that was right for the line and wrong for a smile. Two edges, two masters: the top edge of the beard belongs to the nose and stays rigid; the bottom edge belongs to the mouth and partly nestles against it (0.4) – a smile lifts the wings, a straight line flattens them, the sweep of the sketch remains. On top of that the wings shear upward at the corners: not at all at the root, at the tip just far enough to stay above the corner. On the crooked mouth only one wing tilts.

The moustache is drawn in the mouth's field, offset upward, not in a field of its own above it: lip and beard edge go through the same mapping and foreshorten alike when the head turns. And it takes its measure from the resting shape, not from the babbling mouth – speaking is the lower lip's business; a moustache does not bounce along.

The beard also lies under the nose, in the drawing order too: whatever overlaps the nose, it covers.

Two beard categories

Moustache and full beard are two categories, not one. The moustache lives between the tip of the nose and the upper lip, the full beard on cheek, jaw and chin – two zones, two draws, and with that every combination instead of a list of ready-made pairs. How the full beard is built is in the chapter Hair and hat; it is geometry and belongs with the masses.

Drawn separately does not mean drawn independently. Because vollbart is drawn after bart, it may look at it, and three couplings do:

A fourth coupling does not point at the moustache but at the hairstyle: how the full beard's edge runs out depends on the edge of the hair. A toothed hairline gives the beard an angular edge and its number of teeth, a curl cloud gives it round lobes, a braid a trimmed edge. Hairstyles without a mass edge – trimmings, caps, bald – get a graphic beard. It is not in the catalogue because it forbids nothing: it changes no feature, only its handwriting. The table is in the chapter Hair and hat.

Collars and ears

Five variants, and they fall into two kinds. rund (weight 2.6) and v (0.6) are a hint: a line, not a body, only the proof that the head sits on something. hemd (1.5) and wolle (1.3) are a garment – with cloth, an edge and a back behind the nape. keiner sits at 1.5.

All five are the same ring in body space (styles/klar/kragen.ts): a horizontal ellipse below the chin, parameterised by the angle φ – 0 front centre, ±π at the back. Whatever sits on the collar sits at some φ. Three things then come out for free: the neckline dips at the front (all ring points are at the same height, but the front one has z > 0 and hence a larger s = F/(F−z)), the ribs compress towards the sides, and behind is behind – the rear piece is drawn before the neck's skin and hidden by it.

The ring follows the rotation damped: yaw at 0.3, pitch at 0.15, roll at 0.32. A body turns with the head, but less; undamped, the neck would stop being a neck and become a plinth. Plus a small angle of its own per head – the collar sits crooked.

The shirt collar has a stand at the neckline, broken by the opening at the front, and two leaves on it: top edge at the neckline, outer corner barely lower, and from there the long lower edge running inwards to the point beside the middle of the neck. The knitted hem is a band with a profile – the roll of the knitted edge at the top, almost straight below it, barely opening towards the bottom – with seven to twelve ribs, and every fourth piece of knitting drops one. It may grow tall enough to take up half the neck, and it grows UPWARDS: its lower edge is fixed, so the collar shortens the neck instead of stacking underneath it.

The neck's skin runs down to the hem: its lower edge is the collar's top edge, clipped to the neck's width – before, it ended flat above the collar and left paper in between. How wide the neck is relative to the chin varies per head (head data halsBreite, 0.75 to 1.45): a stalk or a bull neck. The ring takes its measure from exactly that width.

The lampshade

Symptom: A flat disc below the chin with a stalk poking through its middle. Not a collar, a piece of furniture.

Cause: The ring took its measure from the HEAD, not from the neck – three times too wide. A collar is not a disc but a tube: it is not the width that makes it, but the height by which the neck disappears into it.

Fix: A ring just over neck width. The width comes from what hangs on it – the roll of the knitting, the leaves of the shirt. And the roll belongs at the TOP: in the middle of the band's height it turns into a life ring.

The second collar under the collar

Symptom: A darker shadow sticks out below the finished collar, coming from nowhere.

Cause: Under every part lies ONE area, the ground – without it, paper flashes between band, leaves and back, because every area wobbles on its own and shifts with the misregistration. If that ground reaches deeper than what lies in front of it, it peeks out at the bottom.

Fix: Its lower edge follows the shape instead of being flat – on the shirt deep in the middle, where the points hang, shallow at the sides, where the stand ends.

The ear lifts off the head along the normal, not in the surface: across the surface there is barely any room at an anchor near the edge (fx goes towards zero), and the concha folded into the cheek.

The fallen-off handle

Symptom: A gap yawns between ear and head. The ear floats beside it.

Cause: The ear was attached flush to the skin. But the drawn contour is a smoothed silhouette and lies a little inside the point where the ear sits.

Fix: The surface reaches further into the head than the line does – a seam behind the root. Flush is not enough: the edge of the surface trembles (wackel), and at the join the paper showed through.

And the ear has no line at its root. Instead of a second edge beside the contour, the ends of its arc run a little way over it onto the cheek and stop there – so the near ear lies on the head and not beside it. How far is decided by the rotation: by the inset, that is by exactly the amount the anchor has turned behind the silhouette. Frontally that is zero, and the ear sits at the edge.

Measurement is taken against the drawn outline, point by point – not at a fixed depth below the anchor. The head tapers above and below the ear; measured at a fixed depth the upper end would stand deep in the face and the lower one in mid-air. Only the line is clipped in the process: the surface reaches further in, or the paper shows through at the join. If the ear lies at the back, the head surface erases the overrun anyway, and the contour is its edge.

The bitten ear

Symptom: On a turned head the near ear shrinks to a sickle at the edge, and the contour runs straight through it.

Cause: Both ears lay under the head surface. That is right for the far one – but the near one travels behind the silhouette as the head turns (at full rotation a twentieth of the head width) and then lies in front of the cheek, not at its edge.

Fix: A layer of its own, zugewandtes-ohr, behind the head surface. Who lands there is decided by the normal of the ear anchor – a statement in space, not a side in the picture. See Layers.

Trimmings

The earring hangs vertically in the picture, not in head space: gravity does not turn with the head. Its foot sits at the ear, the ring itself is hung below it in picture coordinates.

The plaster is a cross of two strips on one cheek, slightly slanted. The strips are lighter than the skin, with an ink border – a plaster is an overlay, not a hole.

Glasses

Three variants: none, round, square. The reasoning for their construction is in the chapter Space – in short: the glasses are rigid. They are not drawn through the field, which hugs the curvature, but as a real figure in head space: two circles in the frontal plane, a bridge, two temples to the ear, rotated and projected like the skull itself.

Learned in the first press: lenses that foreshorten like skin look painted on. A turned head foreshortens a lens only by the cosine of the rotation – and that is exactly what a lens does.

The tint

A lens is not empty. Every head rolls a colour and a strength for it (wuerfleGlas, appended at the very end of the head data). The range of the strength starts at zero: clear glass is the normal case for spectacles, and a sheet on which every lens is tinted would be a sunglasses shop. At the top an outlier reaches a proper pair of sunglasses – which still lets the eye through.

The seven colours sit in palette.ts next to the skin and hair tones: smoke, tea, amber, bottle, sky, violet, rosé. Smoke carries almost half of them – a tinted pair of glasses is usually grey, and the coloured lenses should be a find, not a row. They are dark tones with little chroma, and the effect comes from the opacity: a pale lens at high opacity would be frosted glass, and there is nobody behind that.

The tint is drawn as a surface inside the same edge the frame follows – and before it: the ring lies over the lens, otherwise the colour eats a piece out of its rim. Why the surface is printed dry is in the trap below.

The colour plate beside the frame

Symptom: The tint sat visibly beside its ring – a band of colour above the rim at the top, a pale seam at the bottom.

Cause: The press's misregistration is a fixed amount in head space. On the skin it is a breath, on a lens a quarter of the head's width it is a sixteenth of that lens.

Fix: trocken, as with the neck at the chin and the band at the crown – the exception for surfaces that have to meet something. The hand's wobble along the edge stays.

A reflection – the bright band a window mirrors in the glass – was built and taken out again. It does turn the tinted surface into a surface, but it lies across the eye; a sheet of sixteen heads became restless with it. Without it the lens reads as a filter rather than a pane, and that is the quieter mistake.