How to draw a box in perspective

If you want it as a course

5 stages, each one taking a support away

Optional, and it changes nothing about how the measurement works — a stage decides which supports are on the sheet and nothing else. You move on when the measurement says you have it rather than after a fixed number of drawings, and skipping ahead or stepping back is always available. Leave it and the page is just a page with a drill on it.

1. 01

   Built with you

   The construction, one mark at a time, with every mark checked before the next.

   clears when you finish the build

2. 02

   Points on the page

   The horizon and the vanishing points are there; the twelve edges are yours. Nothing is drawn for you.

   clears on 2 attempts at 78 or better

3. 03

   You place them

   Put the points down yourself — the horizon is given, where they sit along it is yours. They vanish, you draw, and then you find out where your edges really agreed.

   clears on 2 attempts at 78 or better

4. 04

   Tipped over

   The same points, on a view that is tipped. Nothing is vertical and nothing is level, so no edge can be got right by drawing it straight up — which is what a freely rotated box asks of you.

   clears on 2 attempts at 78 or better

5. 05

   No points at all

   An empty sheet. You decide where the points are and hold them in your head, which is the skill the whole exercise is for.

   clears on 3 attempts at 80 or better

6. ✓

   When you have got it

   Three boxes in a row above 85 with nothing named, drawn at different angles rather than the same comfortable one. Then the real test: draw one on paper and extend the lines yourself. If you can predict what the extensions will show before you draw them, you have internalised the thing the 250 box challenge is trying to teach, and you can stop counting.

Start at stage 1 · Built with youtakes you to the instrument and builds one with you

What it can name

The faults, drawn

A percentage on its own tells you nothing you did not already know. These are the separable problems the measurement identifies — each with a different fix, which is the whole reason for telling them apart. Every example below is a real stroke, generated by the same code that tests the engine.

• Loose aim

  Each edge aimed on its own rather than at a shared point. Every line looks plausible and no two of them describe the same space.

• No structure

  The edges do not sort into three directions at all — built corner to corner, with each line drawn to meet the last rather than to point somewhere.

• Flattened

  Two of the three directions too close together, so the form reads as a panel rather than a box. Usually drawn out from a front face.

• Over-converged

  Vanishing points far too close, which inflates the near corner. The usual overcorrection right after learning that edges converge at all.

• Wobbling edges

  A separate problem from aim: lines drawn slowly get steered, and every correction the hand cannot feel lands on the paper.

• Not drawn through

  The three hidden edges left out. The back corner is where accumulated error shows, so leaving it off hides the thing worth seeing.

How the number is made

What the score actually measures

Nothing here is compared against a reference box, because there is no such thing as the correct box — there are only boxes whose edges agree with each other and boxes whose edges do not. Every number is a relationship inside your own drawing, which is why any box, at any angle, in any style, is measured the same way. Every threshold, and the evidence behind it.

One-point, two-point, three-point: what actually differs

A box in perspective is three sets of four parallel edges, and the number of vanishing points is simply how many of those three sets are heading away from you. Square on to the front face, only the depth edges recede, and they meet at one point. Looking at a corner instead, two sets recede to their own points and the uprights stay vertical: two points. Looking up or down as well, nothing stays parallel and the uprights converge too: three points. Everything else about drawing one follows from that.

These are not three difficulty levels. They are three positions you can be standing in relative to the box, and each one leaves a different number of edge sets running parallel. That is the whole distinction, and once you see it that way the names stop being jargon.

A box has three sets of parallel edges — three directions, four edges each. Perspective is what happens to a set of parallel edges as it recedes: it converges. A set that is NOT receding, because it runs across your view rather than away from it, stays parallel. So the question "how many points?" is really the question "how many of my three directions are heading away from me?"

ONE-POINT. You are square on to the front face, like standing in front of a doorway. The face nearest you is not receding at all, so its uprights stay vertical and its level edges stay horizontal — it is a plain rectangle, drawn with no perspective in it whatsoever. Only the depth edges run away from you, and they all converge on a single point, usually somewhere near the middle of the picture. Draw it by starting with that rectangle, putting one dot where you want the point, and running every depth edge at that dot. It is the easiest to get right and the one most beginners meet first, which is why this drill opens on it.

TWO-POINT. Step round so you are looking at a corner rather than a face. Now two sets are receding — the ones going away to the left and away to the right — and each converges on its own point, both usually well off the page. The uprights are the odd set out: you are level with the box, neither above nor below it, so they are not receding at all and stay vertical. This is the standard view of a building from across a street, and it is what most people mean when they say "in perspective". The single most common mistake is letting the uprights lean, which quietly turns it into a bad three-point drawing.

THREE-POINT. Now look up at the box, or down on it. The uprights are receding too, so they converge as well, on a third point far above or far below. Nothing stays parallel. This is the view up the side of a tower or down onto a crate at your feet, and it is also, roughly, what freehand boxes drawn without a ruler tend to be — which is why the 250 box challenge is effectively a three-point exercise even though nobody draws the points.

The mode you pick above the canvas is not cosmetic: it changes what counts as correct. Uprights that converge are right in three-point and a fault in two-point, and this page will say so, because a measurement that does not know which exercise you meant is guessing.

The check you are currently doing with a ruler

If you are working through the 250 box challenge you already know this routine: draw a box, then extend all twelve lines back with a ruler and see whether each set of four meets in one place. It is the part of the exercise that actually teaches you something, and it is slow, and it is the part people skip when they are tired.

It is also pure arithmetic. Each set of four edges either agrees on a single point or it does not, and the amount by which it does not is an angle. There is nothing to interpret, nothing to have an opinion about, and no reason it should cost anything. This page does it the moment you lift your pen: it groups your edges into three sets by direction, fits the vanishing point that best explains each set, and tells you in degrees how far each set is from agreeing — and which set is the worst.

What it will not do is tell you a box is wrong when it cannot see that it is. If your three directions are too close together, or your edges scattered enough that grouping them is guesswork, it says so instead of quoting you a precise-looking number computed from a coin flip. A measurement that invents a fault is worse than no measurement at all.

Convergence and linework are different problems

Two things go wrong with a drawn box and they are constantly confused for each other. The first is aim: your edges point at slightly different places, so the form does not describe one consistent space. The second is control: your edges wobble on the way there. They look similar on the page and they have opposite fixes.

A wobbly line is drawn too slowly. The hand makes hundreds of tiny corrections it cannot feel, and every one lands on the paper. The fix is to draw faster and from the shoulder, and it will make your aim temporarily worse — which is fine, because aim is much easier to fix than a shaky hand.

Badly aimed lines are usually drawn without deciding where they are going. The fix is not speed but the pause before the stroke: name the vanishing point, point your arm at it, ghost the movement, then commit. This page reports the two separately so you always know which one you are actually working on.

Why boxes, of all things

A box is the smallest object that has a front, a side and a top — the smallest thing that is genuinely three-dimensional rather than a shape with shading on it. Everything you will ever draw in perspective is a box underneath: a building, a car, a hand, a head. Get the box wrong and no amount of rendering on top of it will read as solid.

It is also the one subject where you can be told exactly how wrong you are. An eye or a cat or a face is partly geometry and mostly judgement, and a measurement of the geometry alone answers a question you did not ask. A box has no judgement in it at all. A box is entirely the question of whether twelve lines agree about three points, so measuring it completely is measuring the thing itself.

That is why this is the drill on this site with the most to say. It is not that boxes matter more than faces. It is that a box is the one thing arithmetic can be honest about all the way down.

Two ways to use this page, and when each one is the right one

LEARN IT builds the box with you. You place the vanishing points and the eye level yourself with sliders and watch the construction change as you move them, then lock the view in and the page asks for one mark at a time: the two uprights, then the level edges, then the depth edges, then the far face. Each mark you draw is checked against the one that was asked for — how far its ends sit from where they should be, how far its direction is off, whether it stopped short or overshot, whether it bowed on the way — and a mark that misses comes off the sheet with the reason named. Nothing is scored, because a line drawn onto a visible target is not a measurement of anything.

MEASURE IT gives you an empty sheet. Draw a box however you like, in whatever stroke order suits you, and every edge is grouped by direction, extended, and fitted to the single vanishing point that best explains its set. What comes back is the angle by which each set fails to agree, which set is worst, and the separable faults behind it.

The order matters. Being walked through a construction teaches you what the marks are; only drawing one unaided measures whether you learned it. So the guided build hands you back an empty sheet when it finishes, and the number you get afterwards is the number that counts. If you want that sequence enforced rather than suggested, the course below does it: five stages, starting with the guided build and ending on a blank page, each one taking a support away and each one cleared by measured attempts rather than by a count of boxes.

What a measured page can do that a video cannot

Every good explanation of perspective on the internet has the same shape: somebody who can already do it draws a box while telling you what they are doing. That is genuinely useful and it stops at the same place every time — the moment you pick up your own pen, you are on your own with the one question that matters, which is whether the box you just drew is actually right. A video cannot answer it. A written tutorial cannot answer it. Only a measurement of your specific drawing can, and that is the part that has always been left to you and a ruler.

Four things on this page need your drawing to exist before they can happen, and none of them can be filmed in advance. The first is the measurement itself: your twelve edges, grouped, extended, and fitted, with the disagreement reported in degrees. The second is per-mark checking during the build, which is a judgement about a line you have this second drawn. The third is PREDICT-THEN-CHECK — you tap down where you intend the vanishing points to be, the points vanish, you draw the box unaided, and afterwards the page fits the point your four edges actually agreed on and tells you how far it is from the one you declared. That is a number about the gap between your intention and your execution, and there is no way to produce it without both.

The fourth is the reprojection: your own edges extended back until they cross, showing the viewpoint your hand was really using rather than the one you meant to use. Drag the point around and the disagreement rises and falls; leave it where the edges agree best and you are looking at where you were actually standing when you drew.

We would rather be specific than superlative, so: those four things are the claim. Not that this teaches perspective better than a good instructor — it does not, and a good instructor will tell you things about drawing that no arithmetic can reach. The claim is narrower and, we think, harder to argue with. If what you want is to find out whether your own box is right, in degrees, on the drawing you just made, with the fault named and the fix attached, this page does more of that than anything else we have been able to find.

Questions