Rate my drawing - free, measured feedback

Why does my drawing look off?

Almost always one of five measurable things, and almost never talent. Those five are line confidence, proportion, placement, whether your shapes actually close, and how much range there is between your lightest and darkest marks. When a drawing looks wrong but nobody can say why, the cause is on that list.

Proportion is the most common single answer. The parts are each fine and they disagree with each other about how big the thing is, which produces a drawing that is subtly wrong everywhere and identifiably wrong nowhere. It is also the most fixable, because it is comparison rather than anything innate.

The second most common is value: an outline with everything inside it at the same weight will always read as flat, however accurate the outline is. The third is line confidence, which does not affect accuracy at all and changes how the whole drawing feels - five short careful passes look uncertain even when they land better than one committed stroke would have.

What gets measured

Five things, all of them technique. Never your creativity, never your style, never whether the subject was a good idea - those are not ours to score, and a tool that scored them would be lying about what it can see.

• Line confidence

  Smooth, committed strokes rather than hesitant, hairy repeats.

• Proportion

  Whether the parts agree with each other and with the reference.

• Placement

  Whether a mark lands where it was aimed.

• Corners and joins

  Whether corners meet and shapes close.

• Value range

  Whether there is a tonal range or everything sits at one weight.

Each one has a drill attached to it, so a low number is never just a number - it comes with the five-minute exercise that moves it.

How it works

Most of the measuring is plain geometry on your actual strokes - how much a line wobbles along its length, whether corners meet, whether a mark landed where it was aimed, how many separate passes went into one line. That is arithmetic, so it is instant, it costs nothing to run, and it cannot make things up. There is no daily limit on it.

Anything geometry cannot see gets a second look from a vision model, and a finding only reaches you if several independent passes agree on it. One pass producing a confident-sounding sentence is exactly the failure mode these tools have, and requiring agreement is the cheapest available defence against it.

You get one thing to work on, not five. Five criticisms at once is how people quit, and it is also less useful - a single change you actually make beats five you read and forget.

Feedback arrives as an observation and a fix, in that order, and never as praise. There is no encouragement layer, because a tool that tells everybody their work is wonderful stops carrying information within about two uses.

What a score does not mean

It is not a verdict on the drawing and it is not a judgement of you. Every number here describes one narrow mechanical property of the marks on the page - a drawing can be excellent with a poor line-confidence score, because a deliberately scratchy style is a choice and the measurement cannot tell a choice from an accident.

We never score creativity, composition as taste, likeness, or how good the idea was. Those are the parts that matter most and they are exactly the parts a machine has no business ranking, so the honest move is to leave them alone rather than to produce a plausible-sounding number about them.

A low number is a next exercise, not a grade. Nothing on this site is gated behind a score, no minimum exists anywhere, and there is no leaderboard for a number to matter on.

What we do when we cannot measure something

We show nothing. Not a zero, not a dash, not a low score with an asterisk - the dimension simply does not appear on your result.

The clearest case is pressure. A mouse and a trackpad report none, and a photograph of paper reports none either, so anything derived from how hard you pressed is unmeasurable on that submission. Scoring it zero would tell you that you had drawn with uniformly light pressure, which is not what happened - it is what happens when the hardware cannot answer the question.

The same rule applies anywhere the geometry is too sparse to compare: a drawing with four strokes in it has nothing to say about proportion, so nothing about proportion is said. Every number you see on a result is one we can stand behind, which is only worth something if the ones we cannot are genuinely absent.

From the score to the fix

Each dimension maps to exactly one drill, and the weakest dimension in your last few drawings is the one you are offered next. That is the whole loop, and it is the half that every other feedback tool leaves out - being told what is wrong is only useful if something follows from it.

Line confidence goes to the straight-line drills. Placement goes to ellipses, which is the hardest possible test of landing a mark where you aimed. Corners and joins go to boxes. Proportion goes to the comparison drill, and value range goes to the value scale.

Line confidence - straight line drills · Placement - ellipse drills · Corners and joins - box drills · Proportion - comparison drills · Value range - value scale exercises

Rating a paper drawing from a phone photo

Paper is a first-class input here rather than a fallback. Photograph the page, upload it, and the same five dimensions are measured - the photo is straightened and cleaned up automatically, so a snapshot taken at an angle on a kitchen table is fine.

For the best result, get even light across the page and avoid your own shadow falling over it, because a shadow is a real change in value and the measurement cannot tell it apart from shading you meant. Fill the frame with the paper and keep the phone roughly parallel to it.

The one difference from drawing here is that a photo has no stroke history: we can see the marks but not the order or the speed they were made in. Anything that depends on that is left unmeasured rather than estimated.

What this is not

We never draw anything for you, we never generate art, and your drawings never train any model. The point of the product is that you get better, which cannot be outsourced - a tool that finishes the drawing for you has solved a different problem and left yours untouched.