TRELLIS.2:Native and Compact Structured Latents for 3D Generation,load single dinov3
In the ./ComfyUI/custom_nodes directory, run the following:
git clone https://github.com/smthemex/ComfyUI_TRELLIS2_SM.git
pip install -r requirements.txt
export TORCH_CUDA_ARCH_LIST="8.9" # dep your cuda list
git clone -b v0.4.0 https://github.com/NVlabs/nvdiffrast.git /tmp/extensions/nvdiffrast
pip install /tmp/extensions/nvdiffrast --no-build-isolation
git clone -b renderutils https://github.com/JeffreyXiang/nvdiffrec.git /tmp/extensions/nvdiffrec
pip install /tmp/extensions/nvdiffrec --no-build-isolation
git clone https://github.com/JeffreyXiang/CuMesh.git /tmp/extensions/CuMesh --recursive
#pip install git+https://github.com/ashawkey/cubvh --no-build-isolation # if false git clone --recursive https://github.com/ashawkey/cubvh
cd cubvh
# edit files ' /cubvh/include/gpu/bvh.cuh ' lines 29, annotation printf // 禁用警告输出,否则在多三角面的生成时 会刷新很多条警告
if (!m_overflowed) {
// printf("WARNING TOO BIG (stack overflow)\n");
}
pip install . --no-build-isolation
# sudo apt install libeigen3-dev #linux
git clone https://github.com/JeffreyXiang/FlexGEMM.git /tmp/extensions/FlexGEMM --recursive
pip install /tmp/extensions/FlexGEMM --no-build-isolation
cd ComfyUI/custom_nodes/ComfyUI_TRELLIS2_SM/TRELLIS_2
mkdir -p /tmp/extensions
cp -r o-voxel /tmp/extensions/o-voxel
pip install /tmp/extensions/o-voxel --no-build-isolation
-- ComfyUI/models/dinov2/
|-- dinov3-vitl16-pretrain-lvd1689m.safetensors # rename or not
|-- RMBG.safetensors # rename or not
-- any_path/TRELLIS.2-4B/
|-- ckpts
|-- all files #所有文件 # if use texture mode need update it 如果使用材质模式,模型的json文件有几个是修改了的,注意同步
|-- pipeline.json
|-- texturing_pipeline.json # texture mode


@article{
xiang2025trellis2,
title={Native and Compact Structured Latents for 3D Generation},
author={Xiang, Jianfeng and Chen, Xiaoxue and Xu, Sicheng and Wang, Ruicheng and Lv, Zelong and Deng, Yu and Zhu, Hongyuan and Dong, Yue and Zhao, Hao and Yuan, Nicholas Jing and Yang, Jiaolong},
journal={Tech report},
year={2025}
}
``
23 commits
Python
82.3%
C++
12.4%
Cuda
4.9%
TRELLIS.2:Native and Compact Structured Latents for 3D Generation,load single dinov3
In the ./ComfyUI/custom_nodes directory, run the following:
git clone https://github.com/smthemex/ComfyUI_TRELLIS2_SM.git
pip install -r requirements.txt
export TORCH_CUDA_ARCH_LIST="8.9" # dep your cuda list
git clone -b v0.4.0 https://github.com/NVlabs/nvdiffrast.git /tmp/extensions/nvdiffrast
pip install /tmp/extensions/nvdiffrast --no-build-isolation
git clone -b renderutils https://github.com/JeffreyXiang/nvdiffrec.git /tmp/extensions/nvdiffrec
pip install /tmp/extensions/nvdiffrec --no-build-isolation
git clone https://github.com/JeffreyXiang/CuMesh.git /tmp/extensions/CuMesh --recursive
#pip install git+https://github.com/ashawkey/cubvh --no-build-isolation # if false git clone --recursive https://github.com/ashawkey/cubvh
cd cubvh
# edit files ' /cubvh/include/gpu/bvh.cuh ' lines 29, annotation printf // 禁用警告输出,否则在多三角面的生成时 会刷新很多条警告
if (!m_overflowed) {
// printf("WARNING TOO BIG (stack overflow)\n");
}
pip install . --no-build-isolation
# sudo apt install libeigen3-dev #linux
git clone https://github.com/JeffreyXiang/FlexGEMM.git /tmp/extensions/FlexGEMM --recursive
pip install /tmp/extensions/FlexGEMM --no-build-isolation
cd ComfyUI/custom_nodes/ComfyUI_TRELLIS2_SM/TRELLIS_2
mkdir -p /tmp/extensions
cp -r o-voxel /tmp/extensions/o-voxel
pip install /tmp/extensions/o-voxel --no-build-isolation
-- ComfyUI/models/dinov2/
|-- dinov3-vitl16-pretrain-lvd1689m.safetensors # rename or not
|-- RMBG.safetensors # rename or not
-- any_path/TRELLIS.2-4B/
|-- ckpts
|-- all files #所有文件 # if use texture mode need update it 如果使用材质模式,模型的json文件有几个是修改了的,注意同步
|-- pipeline.json
|-- texturing_pipeline.json # texture mode


@article{
xiang2025trellis2,
title={Native and Compact Structured Latents for 3D Generation},
author={Xiang, Jianfeng and Chen, Xiaoxue and Xu, Sicheng and Wang, Ruicheng and Lv, Zelong and Deng, Yu and Zhu, Hongyuan and Dong, Yue and Zhao, Hao and Yuan, Nicholas Jing and Yang, Jiaolong},
journal={Tech report},
year={2025}
}
``
23 commits
Python
82.3%
C++
12.4%
Cuda
4.9%