Image to 3D using Hunyuan-3D-2
This is a custom_node for ComfyUI. It converts an image into a 3D file you can import into Blender or whatever 3d software you use. It uses Hunyuan-3D-2 from Tencent.
Make sure you use an image with a transparent background.
This custom node might fail after the first restart. Restart ComfyUI again. Click on the panel ~~☐~~ to look for errors.
You will need to install Visual C++ build tools
If you installed ComfyUI with ComfyUI_windows_portable, it is using a version of python which is missing the libs and header files. The way around it is to either copy some folders or install the module from else where. See issue #
sudo apt install python3-dev libgl-dev
Hunyuan-3D needs g++ 13, Suse has g++ 14+ by default
sudo zypper install g++-13 Mesa-libGL-devel python3-dev
Not recommended because ComfyUI-Manager will auto update when you press the update button. git will need manual updates for every custom_node you have.
cd custom_nodes
git clone https://github.com/niknah/ComfyUI-Hunyuan-3D-2
pip install -r requirements.txt
cd ComfyUI-Hunyuan-3D-2
git submodule update --init # You need to get the submodules if you install from git
Right click on the node, click "convert input to widget", "model".
connect a string node to the "model" input.
Put in the string node the ID of the model on hugging face. ie. tencent/Hunyuan3D-2
The big model files are downloaded into ~/.cache/huggingface/hub