Hunyuan3D 2.0
by Tencent Hunyuan
Tencent's image-to-3D system: a shape model builds the mesh, then a paint model textures it.
AI 3D model generators Open source
1. Price: huggingface.co/tencent/Hunyuan3D-2
Open weights, free to download.
huggingface.co/tencent/Hunyuan3D-2No quote yet: price is read from the official pages.
2. Commercial use: huggingface.co
“THIS LICENSE AGREEMENT DOES NOT APPLY IN THE EUROPEAN UNION, UNITED KINGDOM AND SOUTH KOREA”
Allowed only outside the EU, the UK and South Korea, under 1 million monthly active users, and within the Acceptable Use Policy.
huggingface.co3. Credit the author: huggingface.co
“All distributions to Third Parties (other than through a Hosted Service) must be accompanied by a “Notice” text file”huggingface.co
4. Agent access: github.com/Tencent-Hunyuan/Hunyuan3D-2
There is no package or API. Files are downloaded by hand from the official site.
github.com/Tencent-Hunyuan/Hunyuan3D-2License checked Oct 25, 2026 against 1 official page. Tap a verdict to see its source. How we checked Report an error
Hunyuan3D 2.0: turn one picture into a 3D model
Hunyuan3D 2.0 is a free AI model from Tencent that turns a picture of an object into a 3D model with colors painted on. You run it on your own computer with a strong graphics card, or call a small local service from your app. Its license does not cover the EU, the UK or South Korea, and very large products need a separate deal.
Where you can use it
A 3D object from a photo
Give it one picture of a thing, such as a toy, a chair or a sneaker. You get back its shape as a 3D model to save and open elsewhere.
Paint colors onto a model
Add colors and surface detail to a bare 3D model, either one the model just made or one a person modelled by hand.
A local 3D service for your app
Run it as a small service on your own machine, so your app can send a picture and get a 3D file back.
Make 3D assets inside Blender
Artists can call it from Blender through the official add-on and get the result straight into their scene.
Props for a game prototype Idea
Fill an early game level or a 3D scene with quick stand-in objects made from concept sketches, then replace the best ones later.
3D views on product pages Idea
Turn a clean product photo into a model that shoppers can spin on the product page in a web 3D viewer.
As reported by Tencent Hunyuan: Beats the open and closed 3D generators it was compared with on texture quality and on following the input picture.
Hunyuan3D 2.0 popularity
How to start
Tell your AI assistant
“Set up Hunyuan3D 2.0 from the Tencent-Hunyuan/Hunyuan3D-2 code (hy3dgen 2.0.2) in a virtual environment, generate a shape from product.png with Hunyuan3DDiTFlowMatchingPipeline, paint it with Hunyuan3DPaintPipeline and export product.glb; stop and ask me before downloading the weights.”
Show codeStep 1 or Step 2
Install PyTorch from its official site first, then install the code from the Hunyuan3D-2 folder, including the two texture parts.
pip install -r requirements.txt
pip install -e .
# for texture
cd hy3dgen/texgen/custom_rasterizer
python3 setup.py install
cd ../../..
cd hy3dgen/texgen/differentiable_renderer
python3 setup.py install
Or start the ready-made web page on your own computer, here with the smaller mini model and the low memory mode.
python3 gradio_app.py --model_path tencent/Hunyuan3D-2mini --subfolder hunyuan3d-dit-v2-mini --texgen_model_path tencent/Hunyuan3D-2 --low_vram_mode
Mistakes to avoid
The texture model needs these two compiled parts; the shape model alone works without them.
Shape alone needs about 6 GB, shape plus texture about 16 GB.
The license makes the Notice file a condition of every distribution other than a hosted service.
Use outside the Territory and improving other AI models with outputs are both forbidden by the license.
The docs example listens on every network interface, and the request example sends no key.
License
Tencent Hunyuan 3D 2.0 Community License. Free, also for paid products, but only outside the EU, the UK and South Korea. Products with over 1 million monthly users need a separate license, and outputs may not be used to improve other AI models.
“THIS LICENSE AGREEMENT DOES NOT APPLY IN THE EUROPEAN UNION, UNITED KINGDOM AND SOUTH KOREA” huggingface.co
“All distributions to Third Parties (other than through a Hosted Service) must be accompanied by a “Notice” text file” huggingface.co
“is greater than 1 million monthly active users in the preceding calendar month, You must request a license from Tencent” huggingface.co
Redistribution of the weights is allowed only in the Territory, with the license copy, the Notice file and notes on changed files.
“You must cause any modified files to carry prominent notices stating that You changed the files;” LICENSE on GitHub
AllowedWith conditionsNot allowedNot stated or not checked yet
Is Hunyuan3D 2.0 free for commercial use? The full answer Not legal advice. Check the license page before you ship.
For your agent
License rules tell your agent what it may do with Hunyuan3D 2.0. Paste them into your prompt or AGENTS.md.
- Hunyuan3D 2.0 (AI 3D model generators). Tencent's image-to-3D system: a shape model builds the mesh, then a paint model textures it. Price: free; commercial use with conditions; no credit needed; ship inside your app with conditions.
Connect: download from https://huggingface.co/tencent/Hunyuan3D-2 (no API).
Rule: Not licensed in the EU, the UK or South Korea; do not use, ship or show outputs there.
Rule: Products with over 1 million monthly active users need a separate license from Tencent.
Rule: Follow the Acceptable Use Policy; do not use outputs to improve other AI models.
Rule: Ship the license and the Tencent Hunyuan 3D 2.0 Notice file with redistributed weights.
License: https://huggingface.co/tencent/Hunyuan3D-2/blob/main/LICENSE (checked 2026-10-25; details: https://agentgid.com/assets/hunyuan3d-2/)
Powered by Tencent HunyuanAlternatives
Hunyuan3D 2.0 This page
Pick TRELLIS from Microsoft when you want MIT-licensed weights and outputs as meshes, 3D Gaussians or radiance fields from one image.
Pick TripoSR when speed matters more than detail; its makers cite under 0.5 seconds per model on an A100 graphics card.
Pick Shap-E from OpenAI when you want to start from a text prompt as well as from an image.
Official code and models for generating 3D point clouds from text or images, plus a model that turns point clouds into meshes.
Open-weights vision transformer that predicts depth maps, camera poses and 3D geometry from one or more images.
Questions
Can I use the 3D models it makes in a paid product?
Yes, outside the EU, the UK and South Korea, while your products stay under 1 million monthly active users and you follow the use policy. Tencent claims no rights in the outputs.
Which computers does it run on?
The code supports macOS, Windows and Linux; you need about 6 GB of graphics memory for shapes and 16 GB with colors.
Can I try it without installing anything?
Yes, Tencent runs a web version, Hunyuan3D Studio. Your pictures then go to Tencent instead of staying on your machine.
What file do I get?
The shape comes back as a trimesh object that you save as GLB, OBJ or another 3D format.
Is there a smaller version?
Yes. Hunyuan3D-2mini has a 0.6 billion parameter shape model, next to the 1.1 billion one in the main release.