What problem does it solve? Turning a single reference photo of an object or character into usable 3D geometry normally requires manual modeling or opaque one-shot generators; this Skill rebuilds the subject as code-only procedural Three.js through a staged, verifiable pipeline with visual self-correction. ## Core Features & Use Cases - Staged sculpt pipeline: Blockout, structure, form, material, lighting, interaction, and optimization passes are gated by deterministic scripts, comparison sheets, and AI-vision review scores before code generation proceeds. - Projection-first fidelity: Solve the reference camera, de-light the photo into albedo, and project the image's own pixels onto the mesh for patterned finishes and character likeness instead of approximate procedural materials. - Domain tracks: Hard-surface objects, stylized or likeness-maximized characters with anatomy landmarks, and CS2 knife skins with classification manifests, finish routes, and review gates. - Use Case: Given a photo of a CS2 knife skin, classify the family, author a sculpt spec with a detail inventory, generate a TypeScript Three.js factory, and iterate renders against the reference until silhouette, finish, and hardware gates pass. ## Quick Start Use the img2threejs skill to analyze this reference image and plan a quality-gated procedural Three.js reconstruction of the object it shows.
