ComfyUI 3D Generation Guide: Create 3D Models with Hunyuan3D

What is Hunyuan3D?

Hunyuan3D 2.0 is Tencent's open-source 3D generation model that creates high-fidelity 3D models from images, text descriptions, or sketches. It uses a two-stage process:

1. Geometry generation (Hunyuan3D-DiT) — creates the 3D shape from your input
2. Texture synthesis (Hunyuan3D-Paint) — adds high-resolution (4K) textures to the geometry

The model can generate a complete 3D model in as fast as 30 seconds.

Model Variants

Two Approaches in ComfyUI

Option A: ComfyUI Native Support (Simpler)

ComfyUI now natively supports Hunyuan3D. No extra plugins needed — just update to the latest version. Generates geometry without textures (voxel-style output).

Best for: Quick experiments, users who want simple setup.

Option B: ComfyUI-Hunyuan3DWrapper (Full Featured)

Kijai's wrapper plugin provides complete geometry + texture generation. Requires compiling an additional component.

Best for: Production-quality 3D models with textures.

Option A: Native ComfyUI Workflow

Setup

Update ComfyUI to the latest version. Find the Hunyuan3D workflow templates in Workflows → Browse Templates → 3D.

Model Download

For faster generation, use the turbo version:

• hunyuan3d-dit-v2-mv-turbo

Running the Workflow

1. Load the model in the Image Only Checkpoint Loader node
2. Load your input image(s) — use images with clean backgrounds (white or transparent)
3. Run the workflow
4. Output .glb files are saved to ComfyUI/output/mesh/

Option B: Kijai Wrapper (With Textures)

Plugin Installation

Install via ComfyUI Manager:

• ComfyUI-Hunyuan3DWrapper
• ComfyUI_essentials

Texture Component Compilation

After installing the plugin, you need to compile the texture rendering component. This requires basic familiarity with terminal/command line.

For ComfyUI Desktop:

1. Open the terminal panel in Desktop (toggle button → bottom panel → terminal tab)
2. Navigate to the wrapper directory and install the wheel:

cd custom_nodes/ComfyUI-Hunyuan3DWrapper/wheels
pip install custom_rasterizer-0.1-cp312-cp312-win_amd64.whl

For Portable ComfyUI:

python_embeded\python.exe -m pip install ComfyUI\custom_nodes\ComfyUI-Hunyuan3DWrapper\wheels\custom_rasterizer-0.1-cp312-cp312-win_amd64.whl

Choose the wheel file that matches your Python/CUDA version. If no pre-compiled wheel matches, you'll need to compile from source — see the plugin README for instructions.

Model Download

The texture model and delight model are downloaded automatically by the workflow.

Running the Workflow

1. Load hunyuan3d-dit-v2-0-fp16.safetensors in the Hy3DModelLoader node
2. Load your input image
3. Run — the workflow generates both geometry and textured output
4. Output saved to ComfyUI/output/3D/

Input Image Tips

The quality of your 3D output depends heavily on your input image:

• Clean backgrounds — white or transparent backgrounds produce the best results
• Clear subject — the object should be well-lit and clearly visible
• Single object — one subject per image works best
• Multiple angles (for multi-view) — front, side, and back views improve accuracy

Generating from Text

Hunyuan3D takes image input, not text. To generate 3D from text:

1. Generate an image from your text prompt using any text-to-image workflow
2. Feed that image into the Hunyuan3D workflow

For multi-view input, the ComfyUI-MVAdapter plugin can generate multi-angle images from a single image.

Common Issues and Fixes

custom_rasterizer error (Kijai wrapper)

• The texture component isn't compiled — follow the compilation steps above
• Make sure the wheel file matches your Python and CUDA versions

Output geometry is distorted

• Use a cleaner input image with a white/transparent background
• Ensure the subject fills most of the frame
• Try the multi-view workflow with images from multiple angles

Slow generation

• Use the turbo model variant (hunyuan3d-dit-v2-mv-turbo)
• The fast variant (hunyuan3d-dit-v2-0-fast) halves inference time
• Mini variant (v2-mini, 0.6B parameters) is fastest but lower quality

3D Pack plugin won't install

• ComfyUI 3D Pack has dependency conflicts with the latest ComfyUI
• Use the native workflow or Kijai wrapper instead
• If you need 3D Pack specifically, try Comfy3D-WinPortable

Online Alternatives

If you want to try Hunyuan3D without local setup:

• Hunyuan3D-2 Demo
• Hunyuan3D-2mv Demo
• Hunyuan3D Homepage

Related Guides

• Text to Image — Generate input images for 3D conversion
• Install Custom Nodes — Plugin installation guide