Neural Rigging for blender using RigNet
Blender is an open-source 3D application from the Blender Foundation. RigNet is the Machine Learning prediction for articulated characters.
bRigNet requires SciPy, PyTorch, and torch-geometric, along with torch-scatter and torch-sparse.
Download the Neural Rigging add-on as a .zip file and install it from the blender add-ons window, or copy the code to the blender scripts path
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:32:27_Pacific_Daylight_Time_2019
Cuda compilation tools, release 10.2, V10.2.89
At present, the CUDA toolkit from nVidia is required, it can be found at the manufacturer website
A dependency installer is available in the preferences.
- Install CUDA. At present prebuilt packages support versions 10.1, 10.2, 11.1, 12.6
- In the addon preferences, make sure that the Cuda version is detected correctly.
- Hit the "Install" button. It can take time!
Enable bRigNet in the blender addons, the preferences will show up. Set the Modules path properties to the RigNet environment from the previous step
RigNet requires a trained model. A different location can be set in the addon preferences.
the bRigNet tab will show up in the Viewport tools. Select a character mesh as target. Please make sure it doesn't exceed the 5K triangles. You can use the Decimator modifier to reduce the polycount on a copy of the mesh, and select a Collection of high res model on which to transfer the final weights
Rigs generated using RigNet from the command line can be loaded via the Load Skeleton panel. Please select the _.obj and _.txt file and press the button Load Rignet character
The blender addon doesn't cover training yet. If you want to train your own model, please follow the instructions from the RigNet project.
This addon is released under the GNU General Public License version 3 (GPLv3). The RigNet subfolder is licensed under the General Public License Version 3 (GPLv3), or under a Commercial License.