An agent-driven skill that turns one interior room photo into an editable Blender scene, a Three.js viewer and a portable GLB model.
wesbos/3d-model-anything
An iPhone LiDAR capture app plus Three.js viewer and agent skill for rebuilding rooms as editable 3D scenes.
- Stars
- 2,839
- Forks
- 424
- Open issues
- 3
- Last commit
- 11 days ago
What is it?
3D Model Anything turns a room captured on an iPhone into an editable Three.js scene. The iOS app records LiDAR geometry and photographs, and an AI coding agent uses that evidence to take measurements, build the model, and compare rendered views against the original photos.
The repository bundles the iPhone capture app, a browser-based Three.js viewer, a scan-to-threejs agent skill, and a detailed office reconstruction. The viewer offers dollhouse, walking, overhead, wireframe and knolling views. The workflow is iterative and focused on rooms and interiors rather than producing a finished model in one step. Scanning requires a LiDAR-equipped iPhone or iPad; other devices can import and review exports.
Only the office example is distributed. Original scans and photo collections are excluded, so Scan and Overlay modes are disabled for the demo and only work with your own reference assets.
Key features
- 01Captures LiDAR geometry and photos on iPhone or iPad
- 02Builds editable Three.js scenes from captured room data
- 03Provides dollhouse, walking, overhead, wireframe and knolling viewer modes
- 04Includes an agent skill for measurement, modeling and matched-camera photo comparison
- 05Imports and reviews scan exports on non-LiDAR devices
Use cases
- Recreating a real room as an editable 3D scene for visualization
- Prototyping interior layouts from a physical space with an AI agent
- Exploring a reconstructed office in the browser with the included viewer
In collections
Topics
FAQ
- What is 3d-model-anything?
- An iPhone LiDAR capture app plus Three.js viewer and agent skill for rebuilding rooms as editable 3D scenes. 3D Model Anything turns a room captured on an iPhone into an editable Three.js scene. The iOS app records LiDAR geometry and photographs, and an AI coding agent uses that evidence to take measurements, build the model, and compare rendered views against the original photos.
- Is 3d-model-anything free to use? What license does it have?
- 3d-model-anything is open source under the Other license. Check the license terms before using it in commercial products.
- Is 3d-model-anything still maintained?
- Actively maintained. The last commit to 3d-model-anything was on Oct 27, 2026.
- What can I use 3d-model-anything for?
- Recreating a real room as an editable 3D scene for visualization Prototyping interior layouts from a physical space with an AI agent Exploring a reconstructed office in the browser with the included viewer
- Where can I get 3d-model-anything?
- The source code and installation instructions are on GitHub: https://github.com/wesbos/3d-model-anything
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An Agent Skill that lets Codex, Claude Code or Cursor generate an orbitable Three.js architecture scene that assembles itself brick by brick.
Open-source all-in-one platform for annotating and curating multimodal training data, covering images, 3D LiDAR point clouds, and LLM RLHF.
A general-purpose Structure-from-Motion and Multi-View Stereo pipeline with GUI and command-line interfaces for 3D reconstruction from images.
Open-source node-based visual programming tool for photogrammetry, turning photos into 3D models, camera tracking and HDR images.
Cross-vendor 3D Gaussian Splatting trainer that turns photos or video into splats and textured meshes, running on Vulkan or CUDA without Python.