Block out a 3D character for better prompts

What a block-out is (and is not)

A block-out is a readable massing of primitives — head, torso, limbs, maybe a weapon or prop — arranged so silhouette and relative size are obvious at a glance. It is not final art, not rigged, and not textured. You will still hand the result to an AI image tool, a 3D modeling package, or a concept artist for polish. The block-out's only job is to stop that next step from guessing proportions from adjectives.

Worked example 1: a stocky enemy with a weapon

Goal: an oversized-head, short-legged enemy holding a blunt weapon, ready to attach for an AI game-art pass.

1. Open the whiteboard and switch to 3D.
2. Drag a sphere from the palette for the head. Make it noticeably larger than a realistic head-to-body ratio — exaggeration reads more clearly in a silhouette than a subtle size difference.
3. Drag a cube below it for the torso, and scale it wider than it is tall with R to suggest a stocky build.
4. Drag two cylinders for legs, scaled short — this is the "short legs" trait, communicated by proportion, not a label.
5. Drag two more cylinders for arms. Select one, press W to enter move mode, and hold Space while dragging to duplicate it for the other side symmetrically.
6. Drag a cone or small cylinder into one hand's position for the weapon. Rotate it with E so it reads as being held rather than floating.
7. Right-drag to orbit (with nothing selected) until you find a three-quarter angle where the head, torso width, and weapon are all visible at once — this is usually the angle worth exporting.
8. Optional: select the weapon and change its color so it stands out from the body in the flattened PNG.
9. Download the image and attach it with the character brief below.

Worked example 2: a simple prop (a lantern on a post)

Not every 3D block-out is a character. Props benefit just as much from a quick massing pass when the exact proportions of "post" vs. "lantern" vs. "base" matter to the final asset.

1. Drag a thin, tall cylinder for the post.
2. Drag a wider, short cylinder at the base for a footing.
3. Drag a sphere or small cube near the top for the lantern housing, and scale it down with R until it looks proportionate to the post's thickness.
4. Recolor the lantern housing a distinct color (warm yellow, for example) so it reads as "the light" rather than blending into the post in a single-color render.
5. Orbit to a straight-on view — props usually read fine from directly in front, unlike characters.

This example is deliberately small: three shapes, two minutes. Not every attachment needs the eight-shape complexity of a character. Match the block-out's effort to how much proportion ambiguity actually exists in the object.

What to write alongside the attachment

• Role: "enemy grunt", "player avatar", "NPC vendor", "environment prop".
• Proportions to preserve from the image: "large head, short legs", "thin post, wide base".
• Style: "stylized low-poly", "semi-realistic", "flat-shaded toy-like".
• What the model may invent freely: texture, face detail, clothing trim, surface wear.

Always tell the tool to match the silhouette in the image first, then apply style. Style words given without a block-out tend to drift toward generic outputs; a silhouette anchors the result.

Practical tips that save re-exports

• Start with 5–8 shapes for a character, 2–4 for a prop. More pieces rarely improve the first prompt result.
• Exaggerate distinctive features (a hat, a backpack, an oversized weapon) as separate primitives so they survive PNG compression instead of blending into a parent shape.
• Export two angles if one view hides a feature — a back-mounted backpack, for instance — and attach both images together.
• If scale still gets missed, sketch a quick side-by-side height comparison in 2D (a small figure next to a door-height line) and attach that as a third image.

Where this fits in the workflow

Once the block-out communicates proportion clearly, hand it off following the same recipe as any other sketch — see how to attach a sketch to AI tools. If you are unsure whether a task needs 3D at all versus a flatter 2D diagram, read 2D vs 3D first.