What Happened
StableGen is an open-source Blender add-on (GPL-3.0) that links the most tedious steps of 3D asset creation into one line: model generation → texturing → rigging → export to game engines. v0.5.0 was released on November 4, 2026.
Recently someone recommended it: if you want an Agent to model in Blender, install this first.
What It Can Do
Sign in to unlock this key section
New accounts get 10 free credits, plus 1 more for each daily check-in. Spend 1 credit to unlock and it stays open for you.
The highlights of v0.5.0: you can configure the backend ComfyUI directly in Blender, native support for TRELLIS.2 mesh generation, and LATO.2 to rebuild the model into a low-polygon topology.
How to Run It
StableGen operates inside Blender, and the actual AI computation is handled by ComfyUI (a tool for running image and 3D model generation), which can run on your own machine or on another computer on your local network.
- Blender 4.2 to 4.5, or 5.1 and later. Blender 5.0 is not supported.
- NVIDIA graphics cards are recommended: texturing with SDXL needs about 8GB of VRAM, while Qwen Image needs 16GB or more.
- Windows, Linux, or Macs with Apple chips.
- Model files take up 10GB to 50GB or more, plus about 5GB for the bundled ComfyUI.
- ComfyUI can be installed automatically by the add-on, can use one you already have, or can connect to a remote server. The automatic installation has so far only been tested on Windows.
To install, install the downloaded ZIP from Blender's preferences, then follow the prompts in the StableGen panel in the sidebar to install ComfyUI and the models. See the official Wiki for detailed steps.
Relationship to Agents
StableGen itself is an add-on operated from a Blender panel, not an Agent, and the official README does not describe how an Agent would connect to it. The recommendation above means: before having an Agent do 3D work in Blender, set up this generation, texturing, and rigging pipeline, so the Agent can do more than manual modeling. How to actually let an Agent control Blender requires a separate Blender control tool.
Our Advice
- Run it manually once first, before considering having an Agent operate it.
- Check your graphics card and disk. Without enough VRAM, texturing and generation will be slow or may not run at all, and the model files take up a lot of space.
- Check licenses before commercial use. The add-on itself is GPL-3.0, but the models it calls each have their own licenses, three of which are non-commercial only: the Qwen Image 2.1 series uses the Qwen research license, Bria FIBO 2.0 is CC BY-NC 4.0, and the DINOv3 image encoder used by TRELLIS.2 has its own license. Review each one before using them in a commercial project.
We have not installed and tested it ourselves; the above comes from the project's official documentation.
Comments (0)
No comments yet. Start the conversation.