blockout-model

blockout-model

Purpose

The blockout pass produces a coarse, parametric approximation of the object's primary masses. It is not a finished model — it is a testable hypothesis: "does this bounding envelope match the reference silhouette in all canonical views?" Only after the blockout is plausible does detail work begin.

Inputs

• The // Real Object Brief from prepare-prompt/SKILL.md.
• The reference photo path.

What to do

Step 1 — Declare key params

Map each dimension in the brief's Scale section to a param(). Use the brief's estimated value as the default:

const frameW  = param('frameWidth',  130, { min: 100, max: 160, description: 'outer frame width mm' });
const frameH  = param('frameHeight',  48, { min: 30,  max: 70  });
const bodyD   = param('bodyDepth',     5, { min: 2,   max: 12  });
const bridgeW = param('bridgeWidth',  18, { min: 10,  max: 30  });

All key dimensions must be params, not magic numbers. The blockout must be adjustable after the first canonical-view check without touching geometry code.

Step 2 — Place the reference image overlay

Pick the view plane that shows the most information. For most flat products, the front view is the XZ plane:

referenceImage('./reference.jpg', {
  plane: 'xz',
  anchor: 'origin',
  scale: 'fit-bbox',
  opacity: 0.35,
});

The reference image does not affect the score (it is hidden during scoring). It is for visual alignment during authoring.

Step 3 — Build primary mass boxes

One box() or extrudeRoundedRect() per major component. No holes, no features, no material yet. Translate each into position. For symmetric objects, build one half only and call .mirror() after the blockout check:

// primary lens+frame slab
const frameBody = extrudeRoundedRect(frameW, frameH, bodyD, 4)
  .translate(-frameW / 2, 0, -frameH / 2);

// arm stubs — one side, to be mirrored
const armStub = box(50, armW, armT)
  .translate(frameW / 2, 0, -armT / 2);

const full = frameBody.union(armStub.mirror('YZ'));
return full;