Silhouette before anything else
I spend the first hour of most pieces on a shape nobody will ever see. By the time the drawing is finished it is buried under lines, flats, shading and whatever rendering the deadline allowed. It still decides whether the piece works.
This took an embarrassingly long time to accept. The argument against it is obvious — why refine something that gets covered? The argument for it only became clear once I started keeping the failures instead of deleting them.
The two-second test
Take the sketch, fill it solid black on white, scale it to about a hundred and twenty pixels tall. Look at it for two seconds, no longer. If you cannot tell what the character is doing, the drawing does not work, and no amount of detail will fix it. Detail disappears at that size, and that size is what your work gets viewed at more often than not.
The point of the test is not to produce a good silhouette. It is to fail early. A pose that dies at thumbnail stage costs twenty minutes. The same pose discovered to be unreadable after rendering costs a day, and by then you are too invested to throw it away, which is how bad pieces get finished.
What makes a shape read
Three things, roughly in this order of importance.
Negative space. The gaps matter more than the mass. An arm held against the body vanishes; the same arm with a hand’s width of background showing through becomes legible instantly. When a pose is not working, the fix is usually to open a gap rather than to move the limb further. This is the single most useful thing anyone ever told me about drawing and it took three years to sink in.
Asymmetry of the outer contour. Two sides that mirror each other cancel out — not because symmetry is ugly, but because it carries no information about what the character is doing. A symmetrical standing pose is a character waiting for instructions.
One dominant direction. The eye needs somewhere to go first. Most of my failed poses have two competing diagonals, and the result reads as hesitation rather than action. Cutting one of them is almost always the fix, and it is almost always the one I liked more.
Where this breaks down
Portraits. A bust has almost no silhouette to work with — a head shape and some shoulder. The test says nothing useful and I have wasted time trying to make it apply anyway. For those, the equivalent first-hour problem is the value structure of the face, which is a different note and a harder one.
It also breaks down on characters whose entire appeal is in the markings. A solid black shape of a character defined by an elaborate pattern tells you nothing about why anyone commissioned them. The silhouette still has to work in those cases, but it stops being the interesting problem, and pretending otherwise produces technically correct drawings that miss the point.
The version I would give my past self: the drawing is already decided before you have drawn anything worth looking at.
What I actually do
- Twelve to twenty thumbnails, three minutes each, no undo.
- Fill the promising ones solid, view at 120 px, discard anything ambiguous.
- Usually two survive. Take the one that is harder to draw.
- Only then start the actual sketch.
The fourth point matters more than it looks. Given two poses that both pass, I take the one I am less confident about, because the confident one is the pose I have already drawn forty times and it will come out competent and forgettable. The uncomfortable one occasionally fails outright. It is also the only one that has ever taught me anything.