We use complex objects every day without understanding how they work. UNFOLD makes that hidden engineering explorable.

Upload a photo and Gemma 4 interprets the object, identifies components, and proposes a simplified assembly. UNFOLD transforms that structured response into an interactive 3D blueprint: rotate it, unfold it, select individual parts, and explore their mechanisms and construction.

You can also ask questions. Gemma responds with explanations and validated commands that highlight components or adjust the scene. Three curated examples—a MacBook, desk lamp, and cordless drill—connect shared principles across different objects.

Original work includes the connected learning interface, procedural 3D models, morphing particle introduction, photo-to-blueprint pipeline, geometry validation, and contextual scene controls. Built during the hack day with assistance from OpenAI Codex.

Gemma runs through Google’s Gemini API. Generated assemblies are educational approximations, with visible and inferred components distinguished. The source code is publicly available under the MIT license.