Vi3W
Skip to content

Guides

ComfyUI-3D-Pack: what it does, which 3D models it supports, and how to install it

ComfyUI 3D Pack explained: supported models (Hunyuan3D, TRELLIS, SF3D, TripoSR, InstantMesh), install via Manager, common errors.

2 min read

Three translucent file cards holding a crate, a stone lantern and a rock streaming through a glowing portal
Illustration. One node suite, many 3D models.

Quick answer

What is ComfyUI 3D Pack?

ComfyUI-3D-Pack is an MIT-licensed node suite by MrForExample that adds 3D to ComfyUI: meshes, UV textures, 3D Gaussians and NeRFs, plus image-to-3D models including Hunyuan3D 2.1 and 2.0, TRELLIS, Stable Fast 3D, TripoSR, InstantMesh and Unique3D, each with example workflows. It installs through ComfyUI Manager.

What the 3D Pack adds

Its README describes an extensive node suite that lets ComfyUI process 3D inputs (meshes and UV textures) with algorithms such as 3D Gaussian splatting and NeRF, plus a range of 3D models. In practice:

  • loading, previewing and saving meshes
  • UV and texture tools
  • 3D Gaussian splatting and NeRF utilities
  • wrappers for many image-to-3D models, each with example workflows

Supported models (from the README, November 2026)

  • Hunyuan3D 2.1: full, shape-only and texture-only workflows.
  • Hunyuan3D 2.0: turbo, mini, fast and multiview variants.
  • TRELLIS (Microsoft).
  • Stable Fast 3D (Stability AI).
  • TripoSR, InstantMesh, CRM, Unique3D and more.

Each model keeps its own licence: Hunyuan's excludes the EU, UK and South Korea; Stable Fast 3D's is free under US$1M revenue. See open-source 3D AI licences.

Install

Through ComfyUI Manager (the README's recommended way): search "3D Pack", install, restart.

Manually (Windows portable ComfyUI): from the pack folder, run the embedded Python with -m pip install -r requirements.txt, then install.py. A Docker setup is also documented.

The README's target environment: Windows 10/11, Python 3.12, CUDA 12.4. Other combinations may need you to build extensions yourself.

File cards streaming through a portal
Illustration. The pack's save nodes write GLB and other formats.

First run

  1. Open an example workflow for your chosen model (for example, Hunyuan3D 2.1 Shapegen).
  2. Download that model's weights from the link in the README.
  3. Load an image, queue the workflow, find the GLB in your output folder.

Common errors

  • Build failures for CUDA extensions: your Python, PyTorch and CUDA versions don't match the pack's prebuilt wheels.
  • One model works, another doesn't: each wrapper has its own dependencies; follow its section of the README.
  • Out of memory: use shape-only workflows or lighter models. See GPU for AI 3D generation.

From GLB to Blender

Import with File ‣ Import ‣ glTF 2.0, then reduce and clean: AI 3D model generator for Blender.

Same models, no install

Vi3W runs Hunyuan 3D, TRELLIS and TRELLIS 2 next to commercial models in the browser, retopologises the result, and exports GLB, FBX or OBJ. Try Vi3W.

Frequently asked questions

comfyui 3d pack

Install it from ComfyUI Manager, then load one of its example workflows for the model you want. Its README lists Windows 10/11, Python 3.12 and CUDA 12.4 as the target setup.

comfyui 3d tools

The 3D Pack is the broadest set. Model-specific packs, such as kijai's Hunyuan3D wrapper, are lighter if you only need one model.

Last reviewed