Hunyuan3D-2
Open-weight text-to-3D and image-to-3D with PBR textures in seconds.
About
Hunyuan3D-2 is Tencent's open-weight 3D generation system: a two-stage pipeline where a flow-matching diffusion transformer generates geometry and Hunyuan3D-Paint adds PBR textures (albedo, normal, metallic, roughness), outputting game-ready .glb meshes in about 10 seconds. Turbo variants trade some fidelity for speed. It is among the strongest open options for turning concept art or text prompts into usable 3D assets.
It needs serious GPU (mini ~5GB VRAM, full pipeline ~12GB+), outputs static meshes with no rigging or animation, and the license is Apache 2.0 plus Tencent community terms — verify commercial terms before shipping products. Quality drops on unusual or highly detailed subjects.
What it can do
- Text-to-3D mesh generation
- Image-to-3D conversion
- PBR texture synthesis
- Turbo fast-generation variants
- Gradio UI and Blender integration
Setup
git clone https://github.com/Tencent-Hunyuan/Hunyuan3D-2, pip install -r requirements.txt, and download weights via huggingface-cli (tencent/Hunyuan3D-2, ~15GB total). Or use the Gradio web UI / Blender addon.
Prerequisites
NVIDIA GPU with 12GB+ VRAM (5GB+ for mini), Python 3.10+, PyTorch with CUDA.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 44 (tools-b); Trending score 81/100 from researcher assessment.
★ 5,200 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.