### Environment Setup

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Set up the conda environment and install PyTorch with CUDA support. Ensure the CUDA version matches your machine's configuration.

```shell
# step 1, create conda env
conda create -n hunyuan3d-1 python=3.9 or 3.10 or 3.11 or 3.12
conda activate hunyuan3d-1

# step 2. install torch realated package
which pip # check pip corresponds to python

# modify the cuda version according to your machine (recommended)
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121
```

--------------------------------

### Install Build Tools on Ubuntu

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Install necessary build tools, including GCC and G++, on Ubuntu 22.04 before proceeding with CUDA toolkit installation.

```bash
sudo apt update
sudo apt install gcc g++
```

--------------------------------

### Download and Install CUDA 11.8

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Download the CUDA 11.8 runfile installer for Linux x86_64 and execute it with sudo.

```bash
wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run
sudo sh cuda_11.8.0_520.61.05_linux.run
```

--------------------------------

### Install Hugging Face CLI

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Install the huggingface-cli tool to manage model downloads. Refer to Hugging Face documentation for detailed instructions.

```bash
python3 -m pip install "huggingface_hub[cli]"
```

--------------------------------

### Launch Training

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dpaint/README.md

Start the training process using the provided configuration file and specify a log directory. This command is used for overfitting training tests.

```bash
python3 train.py --base 'cfgs/hunyuan-paint-pbr.yaml' --name overfit --logdir logs/
```

--------------------------------

### Minimal Dataset Configuration Example

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/datasets/README.md

A minimal configuration file for the dataset. Ensure all paths and parameters are correctly set for your data.

```json
[
    {
        "mesh_path": "/path/to/object.glb",
        "surface_path": "/path/to/object.npy",
        "image_path": "/path/to/object.png",
        "num_parts": 4,
        "iou_mean": 0.5,
        "iou_max": 0.9,
        "valid": true
    },
    {
        "...": "..."
    },
    ...
]
```

--------------------------------

### Install Miniconda

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Download the Miniconda3 installer script for Linux x86_64 and execute it. Source the bashrc to activate Miniconda.

```bash
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda-installer.sh
bash miniconda-installer.sh
source ~/.bashrc
```

--------------------------------

### Install PartCrafter Dependencies

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Sets up a conda environment and installs PyTorch with CUDA support, followed by other project dependencies. Ensure you have the correct CUDA version installed.

```bash
conda create -n partcrafter python=3.11.13
conda activate partcrafter
pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu124
```

```bash
git clone https://github.com/wgsxm/PartCrafter.git
cd PartCrafter
bash settings/setup.sh
```

```bash
conda install -c conda-forge libegl libglu pyopengl
```

--------------------------------

### Launch Gradio Demo (Standard)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Start the Gradio web interface for multi-view generation using the standard model. The demo will be accessible at http://0.0.0.0:8080.

```bash
python3 app.py
python3 app.py --save_memory
```

--------------------------------

### Install Optional Acceleration Packages

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Install xformers or flash_attn to potentially accelerate computation. Ensure PyTorch and CUDA versions are compatible.

```bash
pip install xformers --index-url https://download.pytorch.org/whl/cu121
```

```bash
pip install flash_attn
```

--------------------------------

### Install nvidia-container-runtime on Linux

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Install the nvidia-container-runtime package on a Linux host to enable GPU support in Docker.

```bash
sudo apt-get install nvidia-container-runtime
```

--------------------------------

### Configure Color Picker with Coloris

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/web/html/threeVisualizer.html

Initialize Coloris for a color picker element. This example sets up a dark theme with RGB format and predefined swatches.

```javascript
Coloris({ el: '.coloris', theme: 'polaroid', themeMode: 'dark', format: 'rgb', alpha: false, swatches: [ 'rgb(0, 0, 0)', 'rgb(128, 128, 128)', 'rgb(255, 255, 255)', 'rgb(132, 188, 218)', 'rgb(128, 227, 119)', 'rgb(213, 96, 98)' ] });
```

--------------------------------

### Download and Install CUDA 12.1

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Download the CUDA 12.1 runfile installer for Linux x86_64 and execute it with sudo.

```bash
wget https://developer.download.nvidia.com/compute/cuda/12.1.0/local_installers/cuda_12.1.0_530.30.02_linux.run
sudo sh cuda_12.1.0_530.30.02_linux.run
```

--------------------------------

### Launch Gradio Demo (Lite)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Start the Gradio web interface for multi-view generation using the lite model. The demo will be accessible at http://0.0.0.0:8080. Use --save_memory to reduce VRAM usage.

```bash
python3 app.py --use_lite
python3 app.py --use_lite --save_memory
```

--------------------------------

### Manage Multiple CUDA Installations

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Use update-alternatives to manage symbolic links for multiple CUDA installations, setting priorities.

```bash
sudo update-alternatives --install /usr/local/cuda cuda /usr/local/cuda-11.8 10
sudo update-alternatives --install /usr/local/cuda cuda /usr/local/cuda-12.1 10
```

--------------------------------

### Manual Installation of ComfyUI-3D-Pack

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/README.md

Use these bash commands to manually clone the ComfyUI-3D-Pack repository into your ComfyUI custom_nodes directory and install its dependencies. Ensure you are in the correct ComfyUI directory before executing.

```bash
cd Your ComfyUI Root Directory\ComfyUI\custom_nodes\
git clone https://github.com/MrForExample/ComfyUI-3D-Pack.git
cd ComfyUI-3D-Pack

Your ComfyUI Root Directory\python_embeded\python.exe -s -m pip install -r requirements.txt
Your ComfyUI Root Directory\python_embeded\python.exe install.py
```

--------------------------------

### Install Rendering Dependencies

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dshape/tools/README.md

Installs necessary Python packages for Blender's rendering script. Ensure you are using the correct Python executable path for your Blender version.

```bash
/Applications/Blender.app/Contents/Resources/4.1/python/bin/python3.11 -m pip install OpenEXR Imath opencv-python
```

--------------------------------

### Build Wheels with CUDA 11.8 on Windows

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Following the CUDA 11.8 swap, execute the build script for each Python environment to create wheels for this CUDA version. Verify the Python path is accurate for your setup.

```bash
# Example of using the path to the miniconda envs on my PC, you may need to change the python path
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py310_cu118\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py311_cu118\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py312_cu118\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
```

--------------------------------

### Install PyTorch with Specific Versions

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Manually specify PyTorch and torchvision versions for compatibility, especially for Python 3.9. Ensure CUDA version matches your environment.

```bash
# python3.9
pip install torch==2.0.1 torchvision==0.15.2 --index-url https://download.pytorch.org/whl/cu118
```

--------------------------------

### Add CUDA Environment Variables

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Set the PATH and LD_LIBRARY_PATH environment variables for both CUDA 11.8 and CUDA 12.1 installations.

```bash
export PATH="/usr/local/cuda-11.8/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-11.8/lib64:$LD_LIBRARY_PATH"
export PATH="/usr/local/cuda-12.1/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-12.1/lib64:$LD_LIBRARY_PATH"
```

--------------------------------

### Build Wheels with CUDA 12.4 on Windows

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Execute the build script for each Python environment to generate wheels compatible with CUDA 12.4. Ensure the Python path is correctly set for your Miniconda installation.

```bash
# Example of using the path to the miniconda envs on my PC, you may need to change the python path
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py310_cu124\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py311_cu124\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py312_cu124\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
```

--------------------------------

### Execute Blender Rendering Script

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dshape/tools/README.md

Runs the Blender rendering script for converting 3D formats to PLY or rendering condition images. Adjust BLENDER_PATH and INPUT_FILE according to your setup.

```bash
$BLENDER_PATH -b -P render/render.py -- \
    --object ${INPUT_FILE} --geo_mode --resolution 512 \
    --output_folder $OUTPUT_FOLDER
```

--------------------------------

### Generate Image with SD v2.1-base (Zero Terminal SNR)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-2/hunyuan3d-paint-v2-0-turbo/README.md

Use this snippet to generate an image from a text prompt using the SD v2.1-base (Zero Terminal SNR) model. Ensure you have the 'diffusers' library installed and a CUDA-enabled GPU.

```python
from diffusers import StableDiffusionPipeline
prompt = "A young girl smiling"
pipe = StableDiffusionPipeline.from_pretrained("ByteDance/sd2.1-base-zsnr-laionaes6").to("cuda")
pipe(prompt, guidance_scale=7.5, guidance_rescale=0.7).images[0].save("out.jpg")
```

--------------------------------

### Clone Repository

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Begin by cloning the repository to your local machine.

```shell
git clone https://github.com/tencent/Hunyuan3D-1
cd Hunyuan3D-1
```

--------------------------------

### Run Quick Inference

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dpaint/README.md

Execute the `demo.py` script for quick inference using a 3D mesh and a reference image. The result will be saved as `textured_mesh.glb`.

```bash
python3 demo.py
```

--------------------------------

### Run Separated Image-to-3D Pipeline (Lite)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Run the lite image-to-3D pipeline in separate modules for GPUs with at least 10GB of VRAM. Requires specifying the image path and output folder.

```bash
bash scripts/image_to_3d_lite_separately.sh ./demos/example_000.png ./outputs/test # >= 10G
```

--------------------------------

### Run Separated Text-to-3D Pipeline (Lite)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Run the lite text-to-3D pipeline in separate modules for GPUs with at least 14GB of VRAM. Requires specifying the text prompt and output folder.

```bash
bash scripts/text_to_3d_lite_separately.sh 'a lovely rabbit' ./outputs/test # >= 14G
```

--------------------------------

### Train PartCrafter from Scratch

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Train the PartCrafter model using the specified configuration file. Adjust gradient accumulation steps, output directory, and tags as needed.

```bash
bash scripts/train_partcrafter.sh --config configs/mp8_nt512.yaml --use_ema \
  --gradient_accumulation_steps 4 \
  --output_dir output_partcrafter \
  --tag scaleup_mp8_nt512
```

--------------------------------

### Download RealESRGAN Weight

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dpaint/README.md

Manually download the RealESRGAN weight to the `ckpt` folder. This is a prerequisite for quick inference.

```bash
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth -P ckpt
```

--------------------------------

### Download TripoSG Weights

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Download the pre-trained TripoSG model weights from Hugging Face. Store them in the specified directory for training.

```bash
huggingface-cli download VAST-AI/TripoSG --local-dir pretrained_weights/TripoSG
```

--------------------------------

### Initialize Coloris Color Picker

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/web/html/gsVisualizer.html

Initializes the Coloris color picker with specified options. Use this to enable color selection in the UI.

```javascript
Coloris({ el: '.coloris', theme: 'polaroid', themeMode: 'dark', format: 'rgb', alpha: false, swatches: [
    'rgb(0, 0, 0)',
    'rgb(128, 128, 128)',
    'rgb(255, 255, 255)',
    'rgb(132, 188, 218)',
    'rgb(128, 227, 119)',
    'rgb(213, 96, 98)'
  ] });
```

--------------------------------

### Finetune PartCrafter with More Tokens

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Finetune the PartCrafter model with more tokens, which may improve performance. Load pre-trained models and checkpoints, and specify a new tag.

```bash
bash scripts/train_partcrafter.sh --config configs/mp16_nt1024.yaml --use_ema \
  --gradient_accumulation_steps 4 \
  --output_dir output_partcrafter \
  --load_pretrained_model scaleup_mp16_nt512 \
  --load_pretrained_model_ckpt 10 \
  --tag scaleup_mp16_nt1024
```

--------------------------------

### Run Lite Image-to-3D Pipeline

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Execute the lite version of the image-to-3D generation pipeline. Requires approximately 22GB of VRAM, or 18GB with the --save_memory flag.

```bash
bash scripts/image_to_3d_lite.sh
```

--------------------------------

### Build and Run Comfy3D Docker Container

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Build the Comfy3D Docker image and run it, mapping port 8188 and enabling all GPUs.

```bash
docker build -t comfy3d . && docker run --rm -it -p 8188:8188 --gpus all comfy3d
```

--------------------------------

### Run Standard Image-to-3D Pipeline

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Execute the standard image-to-3D generation pipeline using a provided script. Requires approximately 30GB of VRAM, or 24GB with the --save_memory flag.

```bash
bash scripts/image_to_3d_std.sh
```

--------------------------------

### Run Lite Text-to-3D Pipeline

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Execute the lite version of the text-to-3D generation pipeline. Requires approximately 22GB of VRAM, or 18GB with the --save_memory flag.

```bash
bash scripts/text_to_3d_lite.sh
```

--------------------------------

### Finetune PartCrafter with More Parts

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Finetune the PartCrafter model to handle a larger number of parts. Load pre-trained models and checkpoints, and specify a new tag.

```bash
bash scripts/train_partcrafter.sh --config configs/mp16_nt512.yaml --use_ema \
  --gradient_accumulation_steps 4 \
  --output_dir output_partcrafter \
  --load_pretrained_model scaleup_mp8_nt512 \
  --load_pretrained_model_ckpt 10 \
  --tag scaleup_mp16_nt512
```

--------------------------------

### Run Standard Text-to-3D Pipeline

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Execute the standard text-to-3D generation pipeline using a provided script. Requires approximately 30GB of VRAM, or 24GB with the --save_memory flag.

```bash
bash scripts/text_to_3d_std.sh
```

--------------------------------

### Sample Points from Mesh Surface

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/datasets/README.md

Use this script to sample points from the mesh surface of a single 3D object. Ensure the input GLB file and output directory are correctly specified.

```python
python datasets/preprocess/mesh_to_point.py --input assets/objects/scissors.glb --output preprocessed_data
```

--------------------------------

### Build Wheels with CUDA 12.1 on Windows

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

After swapping to CUDA 12.1, run the build script for each Python environment to generate wheels compatible with this CUDA version. Adjust the Python path as necessary.

```bash
# Example of using the path to the miniconda envs on my PC, you may need to change the python path
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py310_cu121\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py311_cu121\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
C:\Users\reall\Softwares\Miniconda\envs\Comy3D_py312_cu121\python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
```

--------------------------------

### Run Separated Image-to-3D Pipeline (Std)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Run the standard image-to-3D pipeline in separate modules for GPUs with at least 16GB of VRAM. Requires specifying the image path and output folder.

```bash
bash scripts/image_to_3d_std_separately.sh ./demos/example_000.png ./outputs/test  # >= 16G
```

--------------------------------

### Run Separated Text-to-3D Pipeline (Std)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Run the standard text-to-3D pipeline in separate modules for GPUs with at least 16GB of VRAM. Requires specifying the text prompt and output folder.

```bash
bash scripts/text_to_3d_std_separately.sh 'a lovely rabbit' ./outputs/test # >= 16G
```

--------------------------------

### Preprocess Folder of Meshes

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/datasets/README.md

Use this script to preprocess an entire folder of meshes. It will process all GLB files in the input directory and generate a configuration file.

```python
python datasets/preprocess/preprocess.py --input assets/objects --output preprocessed_data
```

--------------------------------

### Configure Docker Daemon for NVIDIA Runtime

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Edit or create the /etc/docker/daemon.json file to configure Docker to use the NVIDIA container runtime.

```json
{
    "runtimes": {
        "nvidia": {
            "path": "/usr/bin/nvidia-container-runtime",
            "runtimeArgs": []
        } 
    },
    "default-runtime": "nvidia" 
}
```

--------------------------------

### Build Wheels with CUDA 12.1

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Build wheels for ComfyUI-3D-Pack using Python environments with CUDA 12.1. Ensure CUDA 12.1 is selected via update-alternatives and set CUDA_HOME.

```bash
sudo update-alternatives --config cuda
export CUDA_HOME="/usr/local/cuda-12.1/"
~/miniconda3/envs/Comy3D_py310_cu121/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
~/miniconda3/envs/Comy3D_py311_cu121/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
~/miniconda3/envs/Comy3D_py312_cu121/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
```

--------------------------------

### Download Hunyuan3D-1 Model

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Download the standard Hunyuan3D-1 model for multi-view generation using the Hugging Face CLI. Models are saved to the 'weights' directory.

```bash
mkdir weights
huggingface-cli download tencent/Hunyuan3D-1 --local-dir ./weights
```

--------------------------------

### Run InstantMesh Reconstruction

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Reconstructs a textured 3D mesh from multi-view images using the InstantMesh model. Requires the loaded model, multi-view images, and their camera poses.

```json
{
    "class_type": "[Comfy3D] InstantMesh Reconstruction Model",
    "inputs": {
        "lrm_model": ["Load_InstantMesh_Model", 0],
        "multiview_images": ["Zero123Plus_Model", 0],
        "orbit_camera_poses": ["Zero123Plus_Model", 1],
        "orbit_camera_fovy": 30.0,
        "texture_resolution": 1024
    }
}
```

--------------------------------

### Download Unique3D Models

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/DOCKER_INSTRUCTIONS.md

Download the required models for the Unique3D workflow and place them in the 'models' directory. Ensure you have sufficient disk space.

```bash
wget -c https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter_sd15.safetensors -P models/
wget -c https://huggingface.co/h94/IP-Adapter/resolve/main/models/image_encoder/model.safetensors -P models/
wget -c https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth -P models/
wget -c https://huggingface.co/spaces/Wuvin/Unique3D/resolve/main/ckpt/controlnet-tile/diffusion_pytorch_model.safetensors -P models/
wget -c https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.ckpt -P models/
```

--------------------------------

### Use Built Wheels

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Copy the built wheels folder to the specified ComfyUI custom nodes path and reinstall Comfy3D.

```text
Copy the folder _Wheels_win_py{python_version}_cu{cuda_version} to {Your ComfyUI Root Folder Path}\ComfyUI\custom_nodes\ComfyUI-3D-Pack\_Pre_Builds\_Build_Wheels and then reinstall the Comfy3D
For example, _Wheels_win_py311_cu121 to C:\Users\reall\Softwares\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-3D-Pack\_Pre_Builds\_Build_Wheels
```

--------------------------------

### Load InstantMesh Reconstruction Model

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Loads the InstantMesh reconstruction model, supporting 'instant_mesh_large' for FlexiCubes. This is the first step for mesh reconstruction.

```json
{
    "class_type": "[Comfy3D] Load InstantMesh Reconstruction Model",
    "inputs": {
        "model_name": "instant_mesh_large.ckpt"
    }
}
```

--------------------------------

### Build Wheels with CUDA 11.8

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Build wheels for ComfyUI-3D-Pack using Python environments with CUDA 11.8. Ensure CUDA 11.8 is selected via update-alternatives and set CUDA_HOME.

```bash
sudo update-alternatives --config cuda
export CUDA_HOME="/usr/local/cuda-11.8/"
~/miniconda3/envs/Comy3D_py310_cu118/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
~/miniconda3/envs/Comy3D_py311_cu118/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
~/miniconda3/envs/Comy3D_py312_cu118/bin/python ./ComfyUI-3D-Pack/_Pre_Builds/_Build_Scripts/auto_build_all.py
```

--------------------------------

### Stack Orbit Camera Poses

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Generates camera poses for multi-view rendering and training by defining ranges and steps for orbit radius, elevation, azimuth, and orbit center position.

```python
# ComfyUI Node: [Comfy3D] Stack Orbit Camera Poses
# Input Types:
#   orbit_radius_start/stop/step: FLOAT - camera distance range
#   elevation_start/stop/step: FLOAT - vertical angle (-90 to 90)
#   azimuth_start/stop/step: FLOAT - horizontal angle (-180 to 180)
#   orbit_center_X/Y/Z_start/stop/step: FLOAT - orbit center position
# Output Types:
#   orbit_camposes: ORBIT_CAMPOSES - list of camera poses

```

--------------------------------

### Build Docker Image

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/DOCKER_INSTRUCTIONS.md

Build the Docker image for ComfyUI 3D Pack. The DOCKER_BUILDKIT=0 flag is used to disable buildkit, which is not compatible with the docker-compose file.

```bash
DOCKER_BUILDKIT=0 docker build -t comfy3d --file Dockerfile .
```

```bash
DOCKER_BUILDKIT=0 docker compose build
```

--------------------------------

### Run PartCrafter Inference

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/README.md

Generates a 3D part-level object from an input image. The script automatically downloads necessary model weights. Use the `--rmbg` flag for custom images to remove the background.

```python
python scripts/inference_partcrafter.py \
  --image_path assets/images/np3_2f6ab901c5a84ed6bbdf85a67b22a2ee.png \
  --num_parts 3 --tag robot --render
```

--------------------------------

### Clone ComfyUI-3D-Pack Repository

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/README.md

Clone the ComfyUI-3D-Pack repository from GitHub to your local machine.

```bash
git clone https://github.com/MrForExample/ComfyUI-3D-Pack.git
```

--------------------------------

### Feature Extraction with DINOv2 ViT

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/CharacterGen/3D_Stage/models/base/README.md

Demonstrates how to load the DINOv2 base model and its processor from Hugging Face, process an image, and extract its last hidden states for feature extraction.

```python
from transformers import AutoImageProcessor, AutoModel
from PIL import Image
import requests

url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)

processor = AutoImageProcessor.from_pretrained('facebook/dinov2-base')
model = AutoModel.from_pretrained('facebook/dinov2-base')

inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
last_hidden_states = outputs.last_hidden_state
```

--------------------------------

### Execute Overall Data Processing Pipeline

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/Hunyuan3D_2_1/hy3dshape/tools/README.md

Runs the main data processing pipeline by executing the 'pipeline.sh' script. Ensure the first four variables (INPUT_FILE, OUTPUT_FOLDER, NAME, BLENDER_PATH) are correctly configured in the script.

```bash
bash pipeline.sh
```

--------------------------------

### KUIKIT (kiuikit)

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/_Pre_Builds/_Build_Scripts/dependencies.txt

A utility library, likely for 3D or graphics applications.

```git
https://github.com/ashawkey/kiuikit.git
```

--------------------------------

### OpenGL Coordinate System Visualization

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/README.md

Illustrates the world and camera coordinate systems used in OpenGL, relevant for understanding 3D transformations and rendering. Note the axis orientations and conventions for elevation and azimuth.

```text
      World            Camera        
    
      +y              up  target                                              
      |                |  /
      |                | /
      |______+x        |/______right                                      
      /                /
     /
    /
  +z               forward           

  z-axis is pointing towards you and is coming out of the screen
  elevation: in (-90, 90), from +y to +x is (-90, 0)
  azimuth: in (-180, 180), from +z to +x is (0, 90)
```

--------------------------------

### Run TripoSR Reconstruction

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Performs 3D mesh reconstruction using the loaded TripoSR model. Requires the model, a reference image, and a foreground mask.

```json
{
    "class_type": "[Comfy3D] TripoSR",
    "inputs": {
        "tsr_model": ["Load_TripoSR_Model", 0],
        "reference_image": ["LoadImage", 0],
        "reference_mask": ["RemoveBackground", 0],
        "geometry_extract_resolution": 256,
        "marching_cude_threshold": 25.0
    }
}
```

--------------------------------

### Download HunyuanDiT Model

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-1/README.md

Download the HunyuanDiT v1.1 Distilled model for diffusion-based generation. Models are saved to a specific subdirectory within 'weights'.

```bash
mkdir weights/hunyuanDiT
huggingface-cli download Tencent-Hunyuan/HunyuanDiT-v1.1-Diffusers-Distilled --local-dir ./weights/hunyuanDiT
```

--------------------------------

### Load 3D Mesh Node

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Loads a 3D mesh file from disk. Supports .obj, .ply, and .glb formats with options for resizing, renormalization, and cleaning.

```python
# ComfyUI Node: [Comfy3D] Load 3D Mesh
# Input Types:
#   mesh_file_path: STRING - path to mesh file (relative to input dir or absolute)
#   resize: BOOLEAN - whether to resize the mesh
#   renormal: BOOLEAN - whether to recalculate normals
#   retex: BOOLEAN - whether to reset texture
#   optimizable: BOOLEAN - enable gradient computation
#   clean: BOOLEAN - clean mesh geometry
#   resize_bound: FLOAT - bound for resizing (default: 0.5)
# Output Types:
#   mesh: MESH - loaded mesh object

# Example workflow usage:
{
    "class_type": "[Comfy3D] Load 3D Mesh",
    "inputs": {
        "mesh_file_path": "example_inputs/my_model.glb",
        "resize": false,
        "renormal": true,
        "retex": false,
        "optimizable": false,
        "clean": false,
        "resize_bound": 0.5
    }
}
```

--------------------------------

### Hunyuan3D 2.1 Texture Generation Pipeline

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Loads the Hunyuan3D 2.1 texture generation pipeline and applies PBR materials to a mesh using a generated image. Configure resolution and enable PBR material creation.

```json
# Stage 2: Texture generation with PBR materials
{
    "class_type": "[Comfy3D] Load Hunyuan3D 21 TexGen Pipeline",
    "inputs": {
        "max_num_view": 8,
        "resolution": 768,
        "enable_mmgp": true
    }
}

{
    "class_type": "[Comfy3D] Hunyuan3D 21 TexGen",
    "inputs": {
        "texgen_pipe": ["Load_TexGen", 0],
        "mesh_path": ["Save_Mesh_Path", 0],
        "image": ["ShapeGen_Output", 1],
        "create_pbr": true,
        "use_remesh": false
    }
}
# Outputs textured mesh with albedo, metallic, roughness, and normal maps
```

--------------------------------

### Render Images from 3D Object

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Gen_3D_Modules/PartCrafter/datasets/README.md

This script renders images from a given 3D object. Specify the input GLB file and the output directory for the rendered images.

```python
python datasets/preprocess/render.py --input assets/objects/scissors.glb --output preprocessed_data
```

--------------------------------

### Run Docker Container

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/DOCKER_INSTRUCTIONS.md

Run the ComfyUI 3D Pack Docker container with NVIDIA support and volume mounts for output and models. This command maps local directories to container paths and exposes port 8188.

```bash
docker run -d \
  --name comfy3d \
  --platform linux/amd64 \
  --runtime=nvidia \
  -p 8188:8188 \
  -e NVIDIA_VISIBLE_DEVICES=all \
  -e NVIDIA_DRIVER_CAPABILITIES=all \
  -e COMFYUI_PATH=/app \
  -v $(pwd)/output:/app/output \
  -v $(pwd)/models/diffusion_pytorch_model.safetensors:/app/models/controlnet/control_unique3d_sd15_tile.safetensors \
  -v $(pwd)/models/ip-adapter_sd15.safetensors:/app/models/ipadapter/ip-adapter_sd15.safetensors \
  -v $(pwd)/models/RealESRGAN_x4plus.pth:/app/models/upscale_models/RealESRGAN_x4plus.pth \
  -v $(pwd)/models/v1-5-pruned-emaonly.ckpt:/app/models/checkpoints/v1-5-pruned-emaonly.ckpt \
  -v $(pwd)/models/model.safetensors:/app/models/clip_vision/OpenCLIP-ViT-H-14.safetensors \
  comfy3d
```

```bash
docker compose up
```

--------------------------------

### Load 3D Mesh

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Loads a 3D mesh file from disk. Supports .obj, .ply, and .glb formats with options for resizing, renormalization, and cleaning.

```APIDOC
## Load 3D Mesh

### Description
Loads a 3D mesh file from disk into memory. Supports .obj, .ply, and .glb formats with options for resizing, renormalization, and cleaning the mesh geometry.

### Method
Not applicable (ComfyUI Node)

### Endpoint
Not applicable (ComfyUI Node)

### Parameters
#### Input Types:
- **mesh_file_path** (STRING) - Required - path to mesh file (relative to input dir or absolute)
- **resize** (BOOLEAN) - Optional - whether to resize the mesh
- **renormal** (BOOLEAN) - Optional - whether to recalculate normals
- **retex** (BOOLEAN) - Optional - whether to reset texture
- **optimizable** (BOOLEAN) - Optional - enable gradient computation
- **clean** (BOOLEAN) - Optional - clean mesh geometry
- **resize_bound** (FLOAT) - Optional - bound for resizing (default: 0.5)

#### Output Types:
- **mesh** (MESH) - loaded mesh object

### Request Example
```json
{
    "class_type": "[Comfy3D] Load 3D Mesh",
    "inputs": {
        "mesh_file_path": "example_inputs/my_model.glb",
        "resize": false,
        "renormal": true,
        "retex": false,
        "optimizable": false,
        "clean": false,
        "resize_bound": 0.5
    }
}
```

### Response
#### Success Response (200)
- **mesh** (MESH) - loaded mesh object
```

--------------------------------

### Load TripoSG Image to 3D Model Pipeline

Source: https://context7.com/mrforexample/comfyui-3d-pack/llms.txt

Loads the TripoSG pipeline from VAST-AI for high-quality single-image to 3D reconstruction. Supports flash decoder for faster processing.

```json
# Load pipeline:
{
    "class_type": "[Comfy3D] Load Diffusers Pipeline",
    "inputs": {
        "diffusers_pipeline_name": "TripoSGPipeline",
        "repo_id": "VAST-AI/TripoSG"
    }
}
```

--------------------------------

### Cite Common Diffusion Noise Schedules and Sample Steps are Flawed

Source: https://github.com/mrforexample/comfyui-3d-pack/blob/main/Checkpoints/Diffusers/tencent/Hunyuan3D-2/hunyuan3d-paint-v2-0-turbo/README.md

BibTeX entry for the 'Common Diffusion Noise Schedules and Sample Steps are Flawed' paper.

```bibtex
@misc{lin2023common,
      title={Common Diffusion Noise Schedules and Sample Steps are Flawed},
      author={Shanchuan Lin and Bingchen Liu and Jiashi Li and Xiao Yang},
      year={2023},
      eprint={2305.08891},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
```