Hunyuan3D-2: Transform 2D Images into 3D Models

Introduction to Hunyuan3D-2 Workflow

🎉 7/11/25 Update: Hunyuan3D 2.1 is here! Major upgrade from v2.0 featuring enhanced textures, richer details, and improved lighting effects. Try the new workflow now!

Welcome to Hunyuan3D-2, a cutting-edge ComfyUI workflow designed to effortlessly convert 2D images into stunning 3D models. Leveraging the latest advancements in AI segmentation, neural rendering, and computer vision, Hunyuan3D-2 simplifies the traditionally intricate process of 3D modeling. Whether you're a game developer, 3D artist, or content creator, this workflow allows you to generate high-quality 3D assets complete with textures in just a few clicks.

How Hunyuan3D-2 Works

Hunyuan3D-2 ComfyUI Workflow operates through four main stages:

• Image Preparation: Resizes and removes the background from your input image.
• 3D Mesh Generation: Creates a 3D mesh from the isolated subject.
• Mesh Optimization: Cleans up geometry and reduces polygon count for better performance.
• Texture Application: Renders the model from multiple angles and applies realistic textures using AI-powered texture synthesis.

The final output is a production-ready 3D model in GLB format, suitable for games, AR/VR applications, or further refinement in 3D software.

Using Hunyuan3D-2: A Step-by-Step Guide

Step 1: Load Your Image

Upload your image to the "Load Image" node. For optimal results, use a clear photo of a single subject against a simple background.

Step 2: Execute the Workflow

Click "Queue Execution" to process the entire workflow.

Step 3: Download Your 3D Model

Once processing is complete, find your glb format 3D model in the output folder with the filename "xxx.glb".

Key Features of Hunyuan3D-2

• One-Click 3D Generation: Convert any 2D image into a 3D model with a single execution.
• Automatic Background Removal: Utilizes IoSPyReNet segmentation for clean subject isolation.
• High-Quality Mesh Generation: Advanced VAE decoding for detailed geometry reconstruction.
• Optimized Output: Automatic mesh cleanup and polygon reduction for real-time applications.
• PBR Material Support: Compatible with physically-based rendering pipelines.
• GLB Export Format: Universal format compatible with most 3D applications.
• JIT Acceleration: Just-In-Time compilation for faster processing.
• Reproducible Results: Seed control ensures consistent outputs for production workflows.

Key Nodes and Parameters in Hunyuan3D-2 Workflow

Mesh Generation Nodes

• Hy3D VAE Decode:
  • octree_resolution: 384 (higher resolution yields more details)
  • num_chunks: 32000 (controls mesh detail level)
• Hy3D Postprocess Mesh:
  • remove_floaters: true (eliminates disconnected geometry)
  • smooth_normals: false (softens edges when true) <img src="https://cdn.runcomfy.net/workflow_assets/1193/readme01.webp" alt="Hunyuan3D-2" width="450"/>

Texturing Nodes

• Hy3D Sample MultiView:
  • seed: 1024 (controls randomization)
  • denoise_strength: 1.0 (0.7-1.0 recommended for texture clarity)
• CV2 Inpaint Texture:
  • expand_radius: 3 (fills texture gaps; 2-5px recommended) <img src="https://cdn.runcomfy.net/workflow_assets/1193/readme02.webp" alt="Hunyuan3D-2" width="350"/> <img src="https://cdn.runcomfy.net/workflow_assets/1193/readme03.webp" alt="Hunyuan3D-2" width="350"/>

Export Nodes

• Hy3DExportMesh:
  • format: GLB (for universal compatibility) <img src="https://cdn.runcomfy.net/workflow_assets/1193/readme04.webp" alt="Hunyuan3D-2" width="350"/>

Tips for Optimal Results with Hunyuan3D-2

• Use images with clear subjects against simple backgrounds.
• For higher-quality textures, increase the reference image size (requires more VRAM).
• Change the seed value in "Hy3D Sample MultiView" for different texture variations.