Proportion and Shape in Three-Quarter View

Before eyes, hair or a sword, a Game Boy character is a box with a head share, a width and three views. Get those four things right and the details have somewhere to go. Get them wrong and no amount of detail will save the sprite.

What “three-quarter” means for a character

Nintendo’s top-down games — Link’s Awakening, Pokémon Red — are not drawn from directly overhead. The camera is tilted, so you see the front of things and a bit of their top at the same time. For a character that has three consequences:

You see the face and the top of the head. The head is drawn as a full round mass, not a foreshortened dome.
Feet sit under the body, not below it. Legs are short. The bottom row or two is feet, and that is where the character touches the floor.
Lower on screen means nearer. As Raymond Schlitter puts it, “with the top down viewpoint there is implied depth along the Y axis, as we assume objects lower on the screen are closer” (Pixelblog 22). Two characters may overlap vertically and still read; side by side they must not.

Everything below is about the shape that results. Interior shading and features are the next lesson.

Five real characters, measured

Every number in this lesson was counted off these sprites by build-examples.py. Nothing here was drawn for the lesson. Three are 16×24 humanoids by Chasersgaming; two are 16×16 figures by pebonius. All are CC0 — see CREDITS.md.

A four-shade knight sprite facing down, up and sideways, feet on one line. — Down, up and side for each character, cropped to the drawn pixels and stood on one line. Shown at 6×.

A hooded ranger sprite facing down, up and sideways. — Down, up and side for each character, cropped to the drawn pixels and stood on one line. Shown at 6×.

1. The head is a third to a half of the height

Schlitter’s rule for small sprites: “when working with limited size the head usually takes up a third to half of the total sprite.” The sample agrees, and it is worth seeing how it agrees — the 16×24 characters sit at the bottom of that band and the 16×16 characters at the top.

The knight with everything below the head flattened to a neutral. — The front view of each character with every row below the chin recoloured to one flat neutral. Hats and hoods count as head. The chin row is the one judgement call in this lesson; it is pinned in the script so you can argue with it.

The ranger with everything below the hood flattened to a neutral. — The front view of each character with every row below the chin recoloured to one flat neutral. Hats and hoods count as head. The chin row is the one judgement call in this lesson; it is pinned in the script so you can argue with it.

Character	Drawn box	Head rows	Head share	Heads tall
Knight	14×22	7	32%	3.1
Ranger	12×23	9	39%	2.6
Kung Fu Man	12×21	7	33%	3.0
Mini hero	15×16	8	50%	2.0
Mini villager	16×16	9	56%	1.8

Two things to take from the table:

The box decides the proportion. In 24 rows there is room for three heads; in 16 there is room for two. Nobody in the sample fought that. Pick the box first, and the head share follows.
The head is the same height in every direction. Knight 7/7/7, Kung Fu Man 7/7/7, mini hero 8/8/8, ranger 9/8/9 (down/up/side). The head does not shrink when the character turns.

2. Front and back are one shape

In three-quarter view, turning a character round changes what is inside the outline far more than the outline itself.

Knight front silhouette, back silhouette, and a panel marking the few pixels that differ. — Left: the down view as a silhouette. Middle: the up view. Right: only the pixels the two do *not* share, on the faint union. Frames were stood bottom-centred on one canvas before comparing.

Kung fu man front silhouette, back silhouette, and the differing pixels. — Left: the down view as a silhouette. Middle: the up view. Right: only the pixels the two do *not* share, on the faint union. Frames were stood bottom-centred on one canvas before comparing.

Character	Front and back silhouette shared	Front view mirror match
Knight	95.9%	100%
Kung Fu Man	87.9%	100%
Ranger	87.3%	100%
Mini hero	76.8%	76.8%
Mini villager	—	100%

The mini hero is the outlier on both counts for one reason: an asymmetric hat brim. The three humanoids are perfectly mirror-symmetric from the front and keep almost their whole outline when they turn. So the working method is:

Draw the left half of the front view. Mirror it.
Copy the result for the back view. Change the inside — no face, the back of the hair or hood — and touch the outline only where the pose demands (the knight’s back view differs from its front by 8 pixels; the ranger’s by 28).

3. The side view is a narrower column, same height

Character	Front width	Side width	Ratio
Knight	14	7	0.50
Kung Fu Man	12	7	0.58
Ranger	12	9	0.75
Mini hero	15	15	1.00

A humanoid turned sideways is half to three-quarters as wide, and exactly as tall, with the same head rows. The hat, again, is why the mini hero does not narrow: a brim is round from every angle. If your character wears something wide, expect the side view to stay wide.

The side view is also where symmetry stops helping — on the three humanoids the mirror match drops to 69–77% (one arm in front, one foot forward). Draw it once, flip it for the other side.

4. Fill and width

Two shape numbers you already have from earlier lessons still hold here. The front views fill 61–75% of their own box (the course rule is 60–85%), and none has an enclosed hole. And the head is narrower than the body on every 16×24 character:

Character	Head width	Body width	Head ÷ body
Knight	6	14	0.43
Kung Fu Man	8	12	0.67
Ranger	10	12	0.83
Mini hero	14	13	1.08
Mini villager	14	16	0.88

The 16×24 characters get their silhouette from a small head on wide shoulders. The 16×16 characters get theirs from a big head on a small body — hat included. Both read. What does not read is the thing in between: a head the same width as the shoulders, on a 24-row body, which is just a column.

Block in this order, and do not skip ahead.

Box. 16×24 for a hero that stands taller than a tile; 16×16 if you want the Pokémon-era look. Decide once.
Head. A third of the rows in 16×24 (7–9 of 24); half in 16×16 (8–9 of 16). Hats and hoods live inside this band, not on top of it.
Shoulders. Wider than the head in 16×24 — roughly twice as wide on the knight. Narrower than the head in 16×16.
Feet. Legs run to the bottom row. The last one or two rows are feet, no further apart than the shoulders.
Mirror. Draw half, flip it. Then copy it for the back.
Side. Same rows, half to three-quarters of the width.

All of this in Ink only. If the shape does not read as a person as a flat silhouette, the interior will not fix it.

Do it: block in a character, then measure it

Work in Aseprite (or whatever you use) on a 16×24 canvas, Ink only, no interior. Spend no more than ten minutes. Then export the PNG and drop it below.

Block in the front view following the six steps in the card.
Duplicate it, and edit it into the back view. Change as few outline pixels as you can.
Export both as PNG (transparent background if your tool allows; a flat background is fine — the checker clears it).
Load them here. Set the chin row. Read the verdicts.

Then do it again, with the other box. If you drew 16×24, draw the same character at 16×16 and watch the head share jump from a third to a half. Feeling that change with your own pixels is the point; reading it in a table is not.