Home › Vision Models › Hunyuan3D-2.0

Hunyuan3D-2.0 - AI Vision Models Tool

Overview

Hunyuan3D-2.0 is an open-source, production-focused text-to-3D and image-to-3D system from Tencent that generates high-resolution textured meshes using a two-stage diffusion pipeline. The project was publicly released alongside a technical report on January 21, 2025 and includes pretrained shape and texture models plus a user-oriented studio for editing and animating generated assets. ([arxiv.org](https://arxiv.org/abs/2501.12202?utm_source=openai)) The system separates geometry and appearance: Hunyuan3D‑DiT (shape) produces aligned base meshes from a condition (image or text), and Hunyuan3D‑Paint (texture) synthesizes high‑resolution texture maps that can be applied to generated or handcrafted meshes. Tencent positions the release as outperforming prior open- and closed-source 3D generators on geometry detail, condition alignment, and texture quality, and it provides a diffusers-like API plus a web/Gradio demo and Hunyuan3D‑Studio for production workflows. Community reports highlight strong results but note installation steps (custom rasterizer / differentiable renderer) can require extra setup. ([huggingface.co](https://huggingface.co/tencent/Hunyuan3D-2?utm_source=openai))

Model Statistics

  • Likes: 3152

Model Details

Architecture and components: Hunyuan3D-2.0 uses a two-stage pipeline: (1) Hunyuan3D‑DiT — a flow-based diffusion transformer for scalable shape generation that conditions on images/text to produce base meshes, and (2) Hunyuan3D‑Paint — a diffusion-driven texture synthesis model that produces high-resolution UV texture maps for meshes. The project also provides Hunyuan3D‑Studio, a GUI-centered production tool to manipulate, retarget, and animate assets. ([arxiv.org](https://arxiv.org/abs/2501.12202?utm_source=openai)) API and runtime: The authors provide a diffusers-like Python API and a Gradio demo. Typical usage builds meshes with Hunyuan3D‑DiT (returns trimesh objects) and applies textures via Hunyuan3D‑Paint; outputs can be exported to .glb/.obj. The repository includes native components (custom_rasterizer and a differentiable renderer) that must be built on the host system; community threads report that compiling/installing these extensions can be a pain point on some environments. ([huggingface.co](https://huggingface.co/tencent/Hunyuan3D-2?utm_source=openai)) Pretrained artifacts and release: Tencent published pretrained model weights (Hunyuan3D‑DiT‑v2‑0, Hunyuan3D‑Paint‑v2‑0, Hunyuan3D‑Delight‑v2‑0) on Hugging Face and GitHub, and the paper and code are available on arXiv and the project GitHub. Parameter counts and some low-level training hyperparameters are not enumerated in the top-level README and therefore should be consulted in the repo/paper for exact values. ([huggingface.co](https://huggingface.co/tencent/Hunyuan3D-2?utm_source=openai))

Key Features

  • Two-stage pipeline: separate shape (DiT) and texture (Paint) synthesis for higher fidelity.
  • Text-to-3D: generate coherent base meshes directly from text prompts or descriptions.
  • Image-to-3D: single-view or multi-view image conditioning to reconstruct aligned geometry.
  • High-resolution texture maps: production-grade UV painting for generated or custom meshes.
  • Hunyuan3D‑Studio: GUI for editing, retexturing, and animating assets for production workflows.

Example Usage

Example (python):

## Minimal example (from the Hunyuan3D repo)
# install prerequisites per README (PyTorch, custom_rasterizer, differentiable renderer)
from hy3dgen.shapegen import Hunyuan3DDiTFlowMatchingPipeline
from hy3dgen.texgen import Hunyuan3DPaintPipeline

# load shape model and generate a mesh from an image
shape_pipe = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained('tencent/Hunyuan3D-2')
mesh = shape_pipe(image='assets/demo.png')[0]   # returns a trimesh object

# load texture model and paint the mesh
paint_pipe = Hunyuan3DPaintPipeline.from_pretrained('tencent/Hunyuan3D-2')
textured_mesh = paint_pipe(mesh, image='assets/demo.png')

# save as GLB/OBJ using trimesh export
mesh.export('output/model.glb')

# For full examples and advanced options, see the repository README and minimal_demo.py in the official codebase. ([huggingface.co](https://huggingface.co/tencent/Hunyuan3D-2?utm_source=openai))

Pricing

Open-source release: models, weights, and code published by Tencent for public use. No commercial pricing published; the code and pretrained weights are available under the Tencent-Hunyuan-community license and can be run locally or via provided demos. ([github.com](https://github.com/Tencent-Hunyuan/Hunyuan3D-2?utm_source=openai))

Benchmarks

CMMD (lower is better): 3.193 (Source: https://arxiv.org/abs/2501.12202)

FID_CLIP (lower is better): 49.165 (Source: https://arxiv.org/abs/2501.12202)

FID (lower is better): 282.429 (Source: https://arxiv.org/abs/2501.12202)

CLIP-score (higher is better): 0.809 (Source: https://arxiv.org/abs/2501.12202)

Turbo / fast generation claim: Turbo variants: generation in under 30 seconds (Tencent claim for turbo models) (Source: https://www.reuters.com/technology/artificial-intelligence/tencent-expands-ai-push-with-open-source-3d-generation-tools-2025-03-18/)

Last Refreshed: 2026-01-09

Key Information

  • Category: Vision Models
  • Type: AI Vision Models Tool
Visit Official Website

Related Tools

Janus-1.3B

A unified multimodal AI model that decouples visual encoding to support both understanding and generation tasks.
YOLOv10

YOLOv10 is a real-time end-to-end object detection tool that improves upon previous YOLO versions through NMS-free training and a comprehensive architectural design to enhance efficiency and accuracy. It offers state-of-the-art performance across various model sizes and is implemented in PyTorch.
BLIP-2

BLIP-2 is an advanced visual-language model that allows zero-shot image-to-text generation, enabling tasks such as image captioning and visual question answering using a combination of pretrained vision and language models.
DeepSeek-VL2

A series of advanced vision-language models designed for multimodal understanding, available in multiple sizes to suit varying complexity and performance requirements.
YOLOv5

YOLOv5 is a popular open-source AI tool aimed at object detection, image segmentation, and image classification, leveraging PyTorch for model building and deployment. It supports various deployment formats including ONNX, CoreML, and TFLite, and is well-documented for ease of use in research and practical applications.
JanusFlow-1.3B

JanusFlow-1.3B is a unified multimodal model by DeepSeek that integrates autoregressive language models with rectified flow, enabling both multimodal understanding and image generation.