How to Contour a Round Face: Finding the Line on Your Own Face, Not on a Diagram

Which of your five readings a brush can move

This site reads five things off a face: the face length ratio, the forehead ratio, the jaw ratio, the jaw angle and the chin taper. Two of them decide a round result. The face length ratio separates it from the longer outlines, and the jaw angle separates it from the one shape it sits closest to — on a square outline the line below the ear turns a corner you can put a finger on, on a round one it keeps curving and there is no corner to find. The values, the tolerance around each and the weight the classifier gives them are published on the round face guide and on the rule-set page, read at build time out of the same constants the detector runs against. They are deliberately not restated here, because a number copied into prose starts lying the first time the calibration moves and nobody remembers to update the sentence.

Contour moves one of those two and does not move the other. The face length ratio it can move, at both ends: light down the centre column and on the chin point makes the length read longer, shade at the widest part of the cheekbone makes the width read narrower, and since the reading is length over width, either stroke pushes it the same direction. The jaw angle it cannot move. That angle is a property of where the bone turns, and no quantity of pigment relocates a bone. What shade along the jaw actually does is change where the eye guesses the turn is. That is a real effect and a different one, and the difference matters, because the second kind of effect is the one that fails under a light you did not pose for.

The other three readings are movable and mostly not worth moving. Shade at the temples lowers the forehead ratio. Shade along the jaw edge lowers the jaw ratio. Light on the chin point raises the chin taper. But on this outline the forehead and the jaw are already the narrow widths — the cheekbones are the widest point, and that is what makes the shape round — so narrowing either further moves a number that was not deciding anything and spends contrast you have a limited supply of.

Contrast is a budget rather than a tap. Every stroke that lengthens and every stroke that narrows draws on the same pool: how much difference between light and dark a face carries before it reads as makeup instead of as structure. Spend it on the readings that decide the shape and the technique looks effortless. Spend it on the readings that do not and the same amount of product looks like a mistake. That one sentence resolves more of the disagreement in the table below than any individual placement in it does.

Finding the line on your own face

The line has a name in studio work and none in most tutorials: the boundary between the front plane of the face and the side plane, where the surface stops meeting the light and starts turning away from it. It is not the cheekbone and it is not the hollow. On some faces it runs close to one or the other and on many it runs well inside both, which is exactly why a placement inherited from somebody else's face lands in the wrong place on yours.

Yours takes about a minute to find, and the method is deliberately crude, because crude is what makes it repeatable. Finish your base and do not set it. Spray the whole face with water or a setting spray until the skin is visibly wet — the film is what turns the boundary into a change in sheen rather than leaving it as a guess. Dim the room. Hold one light at eye height, straight at your face; a phone torch is enough. Turn your head slowly away from it and watch the shine. At one angle it runs unbroken from the corner of the mouth up past the cheekbone. That band is the boundary. Mark it with a pale liquid while the light is still where it was, because the moment you turn back to the mirror the evidence is gone and you are guessing again.

Two adjustments to the mark. If the face reads wide and you want the front plane narrower, let the line join just past the corner of the mouth. If you want it undeliberate and everyday, join a little further out. Use a pale liquid for this pass rather than a cream stick: a cream is heavy enough to cake in the amount this needs and a liquid is not, a white one shows you the mark while you are making it, and a blue-toned one disappears into the skin once pressed open, which is what you want afterwards and not what you want during.

Three points construct the same line if you would rather build it than watch for it, and all three are findable by touch. The drop point straight below the tail of the brow. The apex of the cheekbone, located with a fingertip run along the bone — not by sucking in your cheeks, which finds a hollow that moves with the expression and drops the bone when you fill it. The corner of the mouth. Above them the turning angle of the frontal tuber at the temple closes the shape; if you cannot feel that one, take a point a finger-width inward from the peak of the brow instead. Join the three lower points with two shallow arcs and the arcs are your boundary, drawn rather than observed.

Then cross-check, and only cross-check. A line dropped straight down from the outer corner of the iris tells you where to stop a stroke aimed at the lip line. Two fingers laid horizontally under the eye tell you roughly how high the cheek stroke should begin. Both catch a mark that has gone obviously wrong. Neither is a measurement: the iris line travels when you tilt your head, and the finger rule changes with the width of the fingers and how hard you press them. Use them to audit the line. Do not use them to draw it.

Eight years of placement advice, side by side

Five published guides for this face shape, spanning September 2018 to April 2026, are worth setting against each other, because they disagree in specific places and agree in others, and the pattern is not the one you would predict going in.

Every date in the first column was read off the page itself, not out of its metadata. Four of the five print a full day — 12 April 2020, 30 May 2025, 8 December 2025 and 7 April 2026 — and are shortened to month and year here only to keep the column even with the September 2018 row, where the page prints no day at all: that date appears once, in the footer tag list. “None given” means the page describes placement without committing to a shade depth. The names are here so that each row can be checked by somebody who is not us. No links are given to any of them, including to the competitor in the last row.

What did not move, and what did

Two things held across all five and eight years. The direction of the cheek stroke is one: every guide draws it from the ear or the hairline inward and downward toward the mouth, and not one draws it the other way or starts at the mouth and works back. The blend direction is the other — inward toward the centre of the face, never straight down. Both are settled. If you take nothing else from the table, take those.

The shade depth also held, which is the stranger result. Where a guide commits to a figure it commits to the same figure: one to two deeper than your skin tone for the shade, one to two lighter for the light, cool or neutral in undertone and nothing that pulls orange. Three of the five commit to no figure at all. So the number is stable and most of the field will not state it, which is a fair description of how much work the number is doing.

The forehead flipped, twice over. September 2018 contours the outer forehead at the hairline and the temples. April 2020 says stay away from the top of the forehead and from the chin, in terms, for this shape specifically. May 2025 does not mention the forehead for a round face. December 2025 puts it back as a C-shaped stroke at the temple blended upward and outward. April 2026 leaves it out again. Four of those five positions are defensible from where each was standing and not one acknowledges the position before it, which is the tell: this is not a technique that improved over eight years. It is an instruction that was never derived from anything, and each guide derived it again from scratch.

The hollow went in and came back out. April 2020 places the shade on the lower edge of the bone and warns against sucking in your cheeks to find it, on the grounds that the hollow you locate that way sits too low and drops the cheekbone. December 2025 tells you to suck in your cheeks to find the hollow and then place the shade just above it. Both are reacting to the same failure and they arrive at opposite instructions, which locates the real disagreement in the finding method rather than in the placement — and the finding method is the part the wet-face pass above replaces outright.

The stopping point drifted from a landmark to a construction to a body part and back to a landmark: where the sideburn and the ear tragus sit; where a vertical from the outer corner of the iris falls; where two fingers under the eye end; the widest part of the cheekbone. Only one of those four is reproducible on a face nobody has seen, and it is the least convenient of the four to apply to your own.

The jaw moved off itself. Three guides put the shade on the jaw — the far edges, tapering toward the chin. December 2025 puts it along the underside of the jawbone and blends downward onto the neck, deliberately, so the head does not read as detached from the body. That is a different line doing a different job, and it is the only change in the table that is an addition rather than a reversal.

The blush rotated ninety degrees. The standing instruction is a diagonal swept upward toward the temple; April 2026 sweeps it vertically downward and claims the same lengthening for it. Both cannot be right about the mechanism. What settles the argument is not the direction but the position — blush laid along the boundary line is the transition layer in the structure below, and blush laid anywhere else is decoration. Direction turns out to be a much smaller variable than five guides treat it as.

Dose, order, and the ceiling on all of it

Begin with the depth the table converges on: one to two deeper than your skin tone, cool or neutral, nothing orange. Then stop relying on it, because depth is not what the depth is for. What you are matching is the colour of the shadow your own face already casts. Look at the underside of your cheekbone and the side of your nose in daylight and match that. On some skins it comes out considerably greyer than one shade deeper and on others considerably less, and the figure is only a starting point either way.

Do not use the applicator that comes with a liquid highlighter. The dose it lays down is several times what this work wants and pressing harder does not reduce it. Transfer to a small detail concealer brush, put the first stroke into the deepest point only, and press it open in place rather than dragging it. Dragging carries the product off the point you identified and onto the fullness beside it, which is the one place it must not go.

Find the deepest point by looking down into a mirror under a single overhead light. That angle puts the hollows — under the eyes, at the base of the nose, along the fold beside the mouth — into their own shadow, and the shadow is the map. Faces hollow in different places and to different degrees. There is no shape anybody can hand you in advance that will be correct for yours, which is why the instructions that come with a diagram are the ones that fail most often.

Three passes and then stop. The first is a detail brush at a normal dose, laid into the deepest point and pressed open. The second picks up less than the first did, and picks it up from the edge of the pan rather than the middle, and drags lightly enough that the skin does not move under the brush. The third changes material instead of adding more: a matte eyeshadow close to your own skin tone, which deposits coverage rather than product and flattens whatever is left without a fourth layer of cream sitting under it. A fourth pass of cream cakes and floats as it sets. More importantly, the light-and-shadow illusion has a ceiling, and that ceiling sits well below the point where you would notice another stroke helping — the last increment is always the one that looks like it should work and is the one that ruins the two before it.

Order differs at the two ends of the line, and the reason is the size of the area. At the cheekbone, pat the outer edge first: the area there is small, so even a deep deposit stays unobtrusive when you turn your head. Along the jaw, pat the inner edge first. The jaw arc is large, and if the deepest product goes onto the outer edge it remains the darkest point at every angle you turn through — a stripe that announces itself the moment you move. That is the failure every guide warns about and none of them explains, and the explanation is an order of operations rather than a quantity.

Cream and liquid carve the range; powder carves the detail. This is a division of labour, not a choice between two menus. A liquid pass with no powder over it has no defined turning point, and the shape stays soft in exactly the place it needed to sharpen. A powder pass with no liquid under it has no surface to sit on and reads as dust across a flat face. Doing one and calling it a shortcut buys half a result, and the half you get is the half nobody notices.

Never let the brightened area touch the contoured area. Leave a strip of your own skin tone running between them. Two zones meeting edge to edge read as a boundary drawn onto the face. A strip of unmodified skin between them reads as a surface turning away from the light, which is the whole claim the technique is making and the only version of it that survives daylight.

Three layers, or a stripe

A face reads small when three values are present on it at once: the brightened centre, the contoured outer ring, and a transition between the two. Any two of the three give you a face with a stripe on it. This is the commonest reason a contour that was placed correctly still looks wrong once you leave the bathroom, and it is invisible in the mirror you placed it by, because that mirror is showing you the layer you just put down rather than the one you left out.

The transition is blush, and it belongs on the boundary line — the same line the wet-face pass found, in a shade from the same tonal family as the contour. A depth-carrying shade laid just inside the outer ring and pressed open with a fingertip in small circles turns the dark into a mid and the mid into skin. Keep the family consistent: if the contour is cool and pink the blush goes deeper in the same pink, and if the contour is warm it goes toward an orange-brown. Two unrelated colours on one face read as a palette rather than as a shadow.

The same product does a second job. A pale shade laid across the inner cheek, the bridge and the root of the nose blends the contour into the blush and puts back the atmosphere that a purely structural pass strips out. Keep it above the level of the nostril. Below that, the mid-face reads as sagging instead of as lifted, and the effect reverses sign while looking almost identical at a glance.

A highlight moved inward carries one constraint. Shifting the cheekbone highlight toward the centre of the face does narrow the outline, but the new position must still overlap the bone's natural prominence. Move it entirely off the bone and the face carries two separate bright points instead of one relocated point — visible at any distance, and it reads as an error rather than as a decision, because there is no structure under the second point to account for it.

Emphasis, not reversal

Contour in everyday makeup emphasises what a face already does. It does not reverse it, and the attempt at reversal is what produces the result people describe as dirty without being able to say why.

The nostril makes the mechanism plain. The instinct for a wide nostril is to shade it, on the reasoning that dark recedes. What happens instead is that a shaded nostril becomes the most contrasted object on the lower half of the face, and contrast is presence: it does not get smaller, it gets more noticeable. The stroke that works is a matte highlight at the transition between the nostril and the mid-face, which takes out the edge separating the two and lets the nostril read as continuous with the cheek. It is not smaller. It has less presence, and those are two different outcomes that look the same from the front and feel nothing like each other from the inside.

The same logic governs a full cheek, and it is the part most often got backwards. The fuller the cheek, the worse heavy contour performs on it, because shading a bulge emphasises the bulge — a sharply defined shadow reads as structure only where there is structure underneath to account for the shadow. On a face carrying a good deal of soft tissue the work is done with light instead: fill the hollows of the mid-face and the fold beside the mouth, and the outer fullness stops reading as though it pushes out past them. The outer contour tightens because the inner one came forward, not because anything out there was shaded.

Which is why the standard advice to contour harder when the first attempt did not show up runs backwards. Past a certain depth more shade stops narrowing the face and starts announcing itself, and the announcement is louder than the narrowing ever was. The ceiling is real and it arrives sooner than the amount of product in your hand suggests.

The rest of the face does the same job

Nose. If the tip is round and full, draw the root slightly wider than you think it ought to be. Root and tip are read against each other, so a narrow root makes a wide tip wider, while a root carrying some width makes the same tip look small by comparison. Do not let the blend pass the front edge of the nostril — beyond that line the root reads as thin, which undoes what you just set out to do. And start the stroke at the inner corner of the eye rather than at the head of the brow: a line beginning at the brow lengthens the nose, and with it the whole mid-face, which is the direction this outline is already working against.

Brow. Deepen along the lower edge of the brow directly above the eyeball and let the tail fade out. A tail drawn heavy makes the side of the face read as turning back sooner than it does, and turning back sooner is what wide means; a tail that disappears does the opposite. Length has a limit as well — the brow should not run past a line taken from the nostril across the cheek. Past it the face reads as all front plane and no side plane, and no quantity of contour puts the side back afterwards.

Lashes. Longest in the middle, not at the outer corner. Outer-corner length extends the eye sideways, and sideways is the axis this outline already has too much of. Middle length opens the eye vertically and adds nothing at all to the width, which is the rare cosmetic change on this shape that costs nothing elsewhere.

Hair. The widest point of the silhouette must not land on the face. That is the entire rule. It is a rule about hair rather than about makeup, and it does more for how the face reads than any stroke on the skin does. If the widest point of your hair currently sits at cheekbone height, widen at the temples until it sits there instead — the same total volume in a different place, and the face is no longer the widest thing in the frame.

Where it stops working

One published account of this technique went back with a ruler after finishing and measured both sides of the face. The contoured side came out marginally wider than the bare one. The photographs looked genuinely different. The tape said they were not.

That is the honest size of the effect, and it is worth stating plainly because everything above is still worth doing. Contour changes where an eye goes and what that eye takes to be the outline of a face. It does not change the outline. The five readings this site measures are taken off the bone and the skin, and a brush reaches neither — which is also why the same face can classify differently on a different day with nobody's technique at fault, and why the result page reports a margin against the neighbouring shapes rather than insisting on a single verdict.

The second limit is light, and it is the one that bites in practice. The wet-face method works because it finds the boundary under one light at one height. Almost none of the lighting you will actually stand in is that: it is a window on one side with a ceiling panel above, or a fitting room with light coming up from below. A placement that reads as structure under a single overhead source can read as a grey smear under a diffuse one. Check the finished face in two or three different lights before deciding whether it worked, and check it from an angle other than straight on, since straight on is the one angle that hides a stripe.

Neither limit argues against doing any of this. Both say what it is. Contour is the cheapest available way to move one of the two readings that decide this shape, it is not a way to move the other one at all, and knowing which of the two you are attempting is most of the skill.