One-, two- and three-point perspective explained by where you are standing rather than by difficulty - plus the checking step the 250 box challenge is built on, why it is the part that actually teaches, and two free tools that do it on the box you just drew.
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 flat shape with shading on it. Everything you will ever draw in perspective is a box underneath: a building, a car, a phone, a hand, a head. Get the box wrong and no amount of rendering on top of it will read as solid.
This is why perspective courses start with boxes, and why the most effective free exercise in learning to draw — Drawabox's 250 box challenge — is 250 of them. It is also why "how do I draw a box in perspective" is a question with a slightly annoying answer: the drawing part is easy to explain in five minutes, and the part that actually makes you better is the part nobody can do for you by explaining.
One-point, two-point, three-point: what actually differs
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 isnot receding, because it runs across your view rather than away from it, stays parallel. So "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. That face is not receding at all, so its uprights stay vertical and its level edges stay horizontal — a plain rectangle with no perspective in it whatsoever. Only the depth edges run away from you, and they converge on a single point. Draw the rectangle, put one dot where you want the point, and run every depth edge at that dot.
Two-point. Step round so you are looking at a corner rather than a face. Now two sets are receding — away to the left and away to the right — each converging on its own point, both usually well off the page. The uprights are the odd set out: you are level with the box, so they are not receding and stay vertical. This is a building seen from across the street, and it is what most people mean by "in perspective". The single commonest 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 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.
This matters more than it looks, because it changes what counts as correct. Uprights that converge are right in three-point and a fault in two-point. Any honest check has to know which one you meant.
Two things go wrong, and they have opposite fixes
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 are constantly confused for each other.
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 far 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 twice without touching down, then commit in one pass.
The checking step is the part that teaches
Here is the bit that gets skipped. In the 250 box challenge you draw a box and then extend all twelve edges back with a ruler to see whether each set of four meets in one place. That extension is not admin — it is the entire lesson. Without it you are just drawing 250 boxes, and drawing 250 of anything without finding out which ones were wrong mostly makes you faster at your current habits.
It gets skipped because it is slow and because it is boring, and both of those are true. 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 and nothing to have an opinion about.
And it is worth being blunt about why you cannot do this step by eye: the judgement that placed the edges is the same judgement being asked to check them. If your sense of "aimed at that point" runs two degrees off, it runs two degrees off when you draw the line and again when you look back at it. That is a general problem with self-assessment, and extending the lines is the classic answer to it: it replaces a judgement with a construction.
What a page can do that an explanation cannot
Everything above is writing. You can read it in four minutes and it will not, by itself, improve a single box — because the question you have after reading it is "is the one I just drew right?", and no article can answer that.
So we built the part that can. The box in perspective drill does the ruler step for you: it groups your edges into three sets by direction, fits the vanishing point that best explains each set, and reports in degrees how far each set is from agreeing — naming which set is worst. Any stroke order you like; the strokes are resolved into edges before anything is measured. Under about one degree is a genuinely well-aimed box.
It will also build one with you, a mark at a time, if you have not got there yet. You place the vanishing points and the eye level with sliders and watch the construction change as you move them, then the page asks for one mark at a time and checks each one as you draw it — how far its ends landed from where they should be, how far its direction is off, whether it stopped short, overshot, or bowed. A mark that misses comes off the sheet with the reason named, so you cannot build the box on top of a bad mark. That part is deliberately not scored: a line drawn onto a visible target is traced, and tracing is fine practice and a worthless measurement. The score comes from the unaided box you draw afterwards.
If you would rather have it as a program than as a tool, there is a five-stage course on the same page. It starts with the guided build and ends on an empty sheet, taking one support away at a time, and you move on when the measurement says you have it rather than after a fixed number of boxes. Two of the middle stages are the ones worth knowing about: in one you tap the vanishing points down yourself, they vanish, you draw the box unaided, and afterwards the page tells you how far the point your edges actually agreed on sits from the one you declared. In another the whole view is tipped over, so no edge can be got right by drawing it straight up — which is what a freely rotated box asks of you, and what most of the 250 box challenge is made of.
Then cylinders, for the same reason
Boxes teach whether your straight edges agree about where they are going. The cylinder in perspective drill asks the same question of your curves, and it is the exercise that follows boxes in every curriculum that teaches boxes at all.
It turns on one fact that most people get backwards. An end ellipse getsmore open the further it is from your eye — and smaller across at the same time. Those two travel in opposite directions as the end recedes, which is exactly why "the far end is wider" is such a confusing way to say it and why we do not say it that way. The far end is the rounder of the two ellipses and the slightly smaller one. A cylinder whose ends sit at the same openness reads as a flat tube with no depth in it.
That page will also build the form with you — centre line, then each end, then the two sides — and it has its own four-stage course. Move the eye level and you can watch both ends change at once, by different amounts, which is the fastest way to stop having to remember the rule.
How to actually use any of this
Draw a box. Extend the lines — on the page or on paper, either is fine. Look at which set was worst and by how much, then draw another one thinking only about that set. Three in a row above 85 with nothing named, at different angles rather than the same comfortable one, and you have got it.
Then do 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.
Everything on both pages is free, unlimited, needs no account, and is computed in your own browser, with no AI model involved at any point. Every threshold and every formula behind the numbers is published at how we measure, including the fixture each threshold was set from, because a score you cannot audit is a score you should not trust.

Tomislav Lombarovic
Founder of Drawize Academy and the creator of the original Drawize game, which has served over 10 million players globally.



