Installing Hunyuan3DWrapper can be tricky, especially if you want to texturize 3D models. Prebuilt rasterizer wheels may fail on certain GPUs, and DLL errors are common. This guide walks you step-by-step through a reliable setup.

https://docs.comfy.org/tutorials/3d/hunyuan3D-2

---

💻 Ready-to-Run Setup Script

Windows (Install_ComfyUI_with_Hunyuan3D-2.bat)

1️⃣ System Prerequisites

Component	Version / Notes
CUDA Toolkit	12.6 (runtime DLLs accessible)
NVIDIA GPU Driver	Latest compatible with CUDA 12.6
Python	3.12.11 (64-bit)
Git	Recent version
Windows	10/11, 64-bit
Visual C++ Redistributable (x64)	2022

⚠️ Missing any of these can lead to DLL load errors.

Make sure you have a ton of free space available in your Disk:

• Python venv: ~6GB
• ComfyUI: ~375MB (without models) ~45GB (with models)

---

2️⃣ Python Virtual Environment

Run in the root directory where python_embeded is located:

python_embeded\python.exe -m venv venv
call venv\Scripts\activate

---

3️⃣ PyTorch 2.6 + CUDA 12.6

Run inside the activated venv, in any directory:

pip install torch==2.6.0+cu126 torchvision==0.21.0+cu126 torchaudio==2.6.0+cu126 --extra-index-url https://download.pytorch.org/whl/cu126

---

4️⃣ Install ComfyUI and Hunyuan3DWrapper

Run in the root directory where ComfyUI folder exists:

git clone https://github.com/comfyanonymous/ComfyUI.git
git clone https://github.com/kijai/ComfyUI-Hunyuan3DWrapper.git ComfyUI\custom_nodes\ComfyUI-Hunyuan3DWrapper

pip install -r ComfyUI\requirements.txt
pip install -r ComfyUI\custom_nodes\ComfyUI-Hunyuan3DWrapper\requirements.txt
pip install transparent_background

:: Optional prebuilt rasterizer wheel
pip install ComfyUI\custom_nodes\ComfyUI-Hunyuan3DWrapper\wheels\custom_rasterizer-0.1.0+torch260.cuda126-cp312-cp312-win_amd64.whl

---

5️⃣ Build Rasterizer from Source (for 3090 / Ampere)

Run inside the folder:

ComfyUI\custom_nodes\ComfyUI-Hunyuan3DWrapper\custom_rasterizer

Steps:

set TORCH_CUDA_ARCH_LIST=8.6
call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
python -m pip install . --force-reinstall

✅ Fixes “no kernel image available” and DLL errors.

---

6️⃣ Launch ComfyUI

Run in the root directory of ComfyUI:

call venv\Scripts\activate
python ComfyUI\main.py --windows-standalone-build

---

7️⃣ 3D Model Workflow

1. Provide Input
   • 2D image (analyzed for depth)
   • Or rough 3D base shape (sphere, cube, cylinder)
2. Generate 3D Model
   • Adjust resolution, mesh detail, textures
   • Click Generate
3. Refine & Texturize
   • Enhance mesh
   • Apply custom textures
4. Export
   • OBJ, FBX, STL
5. Blender Cleanup
   • Merge vertices: M → By Distance
   • Convert triangles to quads: Alt+J
   • Shade Auto Smooth

---

8️⃣ Lessons Learned

• Prebuilt wheels can fail; rebuild from source for Ampere GPUs.
• Match PyTorch, CUDA, Python, and wheel versions.
• Use venv for isolation.
• Ensure MSVC Build Tools + Ninja are installed.

---

TL;DR

1. Install CUDA 12.6, NVIDIA drivers, Git, VC++ redistributable
2. Use a venv
3. Install PyTorch 2.6 + CUDA 12.6, ComfyUI, Hunyuan3DWrapper, transparent_background
4. If rasterizer fails, build from source (TORCH_CUDA_ARCH_LIST=8.6)
5. Launch ComfyUI from venv

---

ComfyUI-BlenderAI-node

Used for AI model generation, next-generation Blender rendering engine, texture enhancement&generation (based on ComfyUI)

Download: https://github.com/AIGODLIKE/ComfyUI-BlenderAI-node