MrForExample/ComfyUI-3D-Pack
FreeAn extensive node suite that enables ComfyUI to process 3D inputs (Mesh & UV Texture, etc) using cutting edge algorithms (3DGS, NeRF, etc.)
FreeFree tier
About MrForExample/ComfyUI-3D-Pack
An extensive node suite for ComfyUI that enables 3D asset generation using cutting-edge algorithms such as 3D Gaussian Splatting (3DGS), Neural Radiance Fields (NeRF), and models like InstantMesh, CRM, and TripoSR. It processes 3D inputs including meshes and UV textures, and includes the PartCrafter pipeline for part segmentation, supporting both single-object and scene modes. The suite is designed to make 3D generation as convenient as image/video generation within ComfyUI, with pre-built installers for Windows, easy installation via ComfyUI-Manager, and automated dependency handling.
Key Features
Integrates with ComfyUI for 3D asset generation
Supports cutting-edge algorithms: 3DGS, NeRF, Marching Cubes
Models: InstantMesh, CRM, TripoSR, and more
PartCrafter pipeline: two-model for part segmentation in single object and scene modes
Outputs merged full mesh ZIP archives with individual part meshes
Pre-built installers for Windows (Python 3.12, CUDA 12.4, torch 2.5.1+cu124)
Automatic pre-build download and installation via install.py
Works with ComfyUI-Manager for easy installation
Includes example workflows and manual download options for pre-trained models
Pros & Cons
Pros
- Open-source and free to use
- Seamless integration with ComfyUI ecosystem
- Supports a wide range of modern 3D algorithms and models
- Part segmentation capability (PartCrafter) for detailed 3D editing
- Pre-built installers simplify setup on Windows
- Active community and regular updates (last update June 2025)
Cons
- Requires specific runtime: Windows 10/11 or Linux, Python 3.12, CUDA 12.4, torch 2.5.1
- Some nodes (InstantNGP conversion) require Visual Studio Build Tools (Windows) or gcc/g++ (Linux)
- Automatic pre-build download may fail in some environments, requiring manual build or manual model downloads
- Not all features are fully automated; some JIT compilation occurs at runtime
- No native support for non-NVIDIA GPUs (CUDA dependency)
Best For
3D mesh generation from a single image with part segmentationScene reconstruction from an image into multiple 3D part meshesProcessing UV textures and meshes with state-of-the-art 3D algorithmsRapid prototyping of 3D assets within ComfyUI workflowsEducational and research purposes in 3D computer vision
FAQ
How do I install ComfyUI-3D-Pack?
It can be installed directly via ComfyUI-Manager, or manually by cloning the repo into ComfyUI/custom_nodes/ and running install.py. Pre-built installers are available for Windows (Python 3.12, CUDA 12.4, torch 2.5.1). For Docker, see DOCKER_INSTRUCTIONS.md.
What are the system requirements?
Windows 10/11 or Linux with Python 3.12, CUDA 12.4, and torch 2.5.1+cu124. Some nodes require Visual Studio Build Tools (Windows) or gcc/g++ (Linux). A compatible NVIDIA GPU is recommended.
How do I troubleshoot missing nodes or dependencies?
Use ComfyUI-Manager to install missing custom nodes. For missing Python libraries, check the requirements.txt and my-reqs.txt files. Pre-trained models can be downloaded manually into the Checkpoints directory without overwriting existing .json files.
Does it support part segmentation?
Yes, via the PartCrafter pipeline. It includes two modes: Single object mode (single image to 3D mesh with part segmentation) and Scene mode (scene image to multiple 3D part meshes). Outputs merged full mesh ZIP archives with individual part meshes.
Is ComfyUI-3D-Pack free?
Yes, it is fully open-source and free to use. There are no paid tiers or subscriptions.