Input Photo
PBR Enabled
Sample Photos (Instant Preview)
Remove Background (rembg)
Synthesize PBR Material (Paint 2.1)
Inference Steps
50
Octree Resolution
256
About Tencent Hunyuan3D-2.1
Hunyuan3D-2.1 is an open-source scalable 3D asset generation system featuring two core technical components:
- Hunyuan3D-Shape-2.1 (3.3B): Flow Matching Diffusion Transformer generating detailed geometry directly from single images.
- Hunyuan3D-Paint-2.1 (2B): Physically-Based Rendering (PBR) texture synthesis predicting albedo, metallic, and roughness maps for cinema-grade illumination.
Community GPU Grant on Hugging Face
If you want to host this model on your own dedicated Hugging Face ZeroGPU or dedicated A100 hardware without recurring fees, you can request a Community GPU Grant in 3 steps:
1. Go to your Space on Hugging Face: https://huggingface.co/spaces/cyber1235896/hunyuan3d-2-1-demo
2. Click on the "Community" tab -> "New Discussion"
3. Title: "Apply for a GPU community grant: Personal project"
4. Explain your open-source 3D generation demo and request ZeroGPU (zero-a10g) hardware!
Python Code Usage (Diffusers-like API)
from hy3dshape.pipelines import Hunyuan3DDiTFlowMatchingPipeline
from hy3dpaint.textureGenPipeline import Hunyuan3DPaintPipeline, Hunyuan3DPaintConfig
# 1. Generate untextured mesh
shape_pipe = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained('tencent/Hunyuan3D-2.1')
mesh_untextured = shape_pipe(image='assets/demo.png')[0]
mesh_untextured.export('demo.glb')
# 2. Synthesize PBR material texture
paint_pipe = Hunyuan3DPaintPipeline(Hunyuan3DPaintConfig(max_num_view=6, resolution=512))
mesh_textured = paint_pipe(mesh_path='demo.glb', image_path='assets/demo.png')