img2threejs
Skill that rebuilds the object in a reference image as a Three.js model, entirely in code. You hand it a photo, you get back a TypeScript factory that draws the object procedurally, ready to animate, running in the browser. The pipeline runs in passes: inventory of visible details, structural validation, then successive refinements until the render holds up. Output is code, not an exported mesh, so you can read and edit what comes out. Written in Python 3.10+, no external dependencies.
Strengths
- Output is readable, versionable code rather than an opaque mesh
- Payload shipped to the browser is nowhere near that of an exported model
- Animation-ready, since every part of the object is a variable
Limitations
- Stalls on organic shapes and rich textures, it needs crisp geometry
- The refinement loop burns tokens, keep it for the pieces that matter
Best for
- A designer who wants an animated 3D object on a page without bloating load time
- Someone prototyping a scene who wants to hand-edit it afterwards
- A front-end dev holding a page weight budget with 3D on the page