Skip to content

Image to 3D Model (ComfyUI + Hunyuan3D 2.1)

Example Image

Special attribution to Joakim Wahlberg from
Fab Lab Halmstad
for introducing this workflow to me.

Installation

Requirements

OS

  • Windows 10 / 11 (64-bit)

GPU

  • NVIDIA GPU
  • Minimum: 8 GB VRAM
  • Recommended: 12–24 GB VRAM

CUDA / PyTorch

  • Latest NVIDIA driver
  • GPU-enabled PyTorch (cu121, cu124, cu128)
  • torch.cuda.is_available() must return True

⚠️ ComfyUI portable includes PyTorch, but it may not be compatible with your GPU. If needed, install a matching CUDA + PyTorch version manually.

Git

  • https://git-scm.com/install/windows

Python

ComfyUI portable ships with embedded Python, no system Python installation required.

  • Python 3.12.x (cp312) ✅
  • ❌ Python 3.13 is not supported (missing wheels: open3d, xatlas)

Required Python Packages

trimesh
pygltflib
pymeshlab
xatlas
open3d
diso   # optional (or set mc_algo="mc")

Setup Steps

1. Install CUDA Toolkit

Download and install CUDA Toolkit for Windows:
https://developer.nvidia.com/cuda-downloads

2. Download ComfyUI Portable

ComfyUI GitHub repository:
https://github.com/Comfy-Org/ComfyUI

Recommended build:
Portable with PyTorch CUDA 12.8 and Python 3.12

Download:
https://github.com/comfyanonymous/ComfyUI/releases/latest/download/ComfyUI_windows_portable_nvidia_cu128.7z

Unzip and move the folder to: 

C:\ComfyUI_windows_portable

3. Install ComfyUI-Manager

ComfyUI-Manager repository:
https://github.com/Comfy-Org/ComfyUI-Manager

  1. Download install-manager-for-portable-version.bat
  2. Place it in: 
    C:\ComfyUI_windows_portable\
  3. Run the script and wait for it to finish

4. Start ComfyUI

Run: 

run_nvidia_gpu.bat

Open in browser: 

http://127.0.0.1:8188/

Install Hunyuan3D 2.1

  1. Open ComfyUI
  2. Open Templates from the left sidebar
  3. Filter by 3D Models
  4. Select Hunyuan3D 2.1
  5. Open the template workflow
  6. Download the model when prompted: 
    hunyuan_3d_v2.1.safetensors
  7. Place it in: 
    C:\ComfyUI\models\checkpoints\
  8. Open ComfyUI Manager and install missing custom nodes
  9. Restart ComfyUI
  10. Upload an image and click Run

Option B — Advanced

Follow manual installation instructions:
https://github.com/visualbruno/ComfyUI-Hunyuan3d-2-1

Download models from:
https://huggingface.co/tencent/Hunyuan3D-2.1

Place files: 

hunyuan3d-dit-v2-1/model.fp16.ckpt
→ C:\ComfyUI_windows_portable\ComfyUI\models\diffusion_models\

hunyuan3d-vae-v2-1/model.fp16.ckpt
→ C:\ComfyUI_windows_portable\ComfyUI\models\vae\

Use: 

run_nvidia_gpu_fast_fp16_accumulation.bat

Workflow examples: https://github.com/visualbruno/ComfyUI-Hunyuan3d-2-1/tree/main/workflow_examples

Copy workflows to: 

C:\ComfyUI_windows_portable\ComfyUI\user\default\workflows