Generate a 3D model of an articulated object, such as a desk lamp or fan, from a plain text description.
Build a large dataset of varied, correctly structured 3D objects for machine learning research.
Fork an existing generated 3D object and modify it with a follow-up instruction.
| mattzh72/articraft | sapientinc/hrm-text | bytedance/lance | |
|---|---|---|---|
| Stars | 621 | 617 | 637 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | 4/5 | 5/5 | 5/5 |
| Audience | researcher | researcher | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires an API key from OpenAI, Gemini, or Anthropic, or an AI coding agent as a substitute.
Articraft is a system for automatically generating 3D models of objects that have moving parts, things like a desk lamp with a hinged arm, a fan with adjustable tilt, or a drawer with sliding components. These are called "articulated" objects because they have joints and movable pieces, which makes them significantly harder to generate than simple static shapes. The system works by using AI language models to write Python code that describes the 3D geometry, the individual semantic parts such as base, arm, and head, and the physical joints connecting them. Instead of a designer manually modeling each object in a 3D tool, you give Articraft a text prompt describing what you want, and it generates the code that produces the 3D asset. You can also fork an existing generated object and modify it with a follow-up instruction like "make the handle longer." The main use case is building large datasets of articulated 3D objects for machine learning research, where researchers need thousands of varied, correctly structured models rather than just a handful of hand-crafted ones. A web viewer lets you browse generated objects and inspect their joints interactively. You connect it to an AI provider API, such as OpenAI, Gemini, or Anthropic, and run generation from the command line. If you do not have API keys, you can also use AI coding agents like Claude Code or Cursor and point them at the repository with a text prompt. Contributions to a shared public dataset are welcomed and released under a Creative Commons license, separate from the code itself. Articraft warns that generated model files are executed as Python code, so only trusted sources should be run. It is written in Python and licensed under Apache 2.0.
Articraft uses AI to automatically generate 3D models of objects with moving parts, like hinges or sliders, by writing code that defines their geometry and joints.
Mainly Python. The stack also includes Python, OpenAI API, Anthropic API.
Apache 2.0: use freely for any purpose, including commercial use, as long as you keep the copyright and license notices.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.