#2828th of 34 in Open modelsFor image, video and design
trellis.2
Image-to-3D generator producing PBR-textured meshes from a 4B-parameter model
Live demo ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 12.4k
- License
- MIT
- Last commit
- Jun 2026
- Language
- Python
4,836 on 31 Mar 202612,355 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
TRELLIS.2 is Microsoft's 4B-parameter image-to-3D model built on a sparse voxel representation called O-Voxel. It generates textured assets at 512³ to 1536³ resolution with base color, roughness, metallic and opacity, and exports GLB. The repo includes inference scripts, a web demo, shape-conditioned texturing, and full training code.
Who it is for: Researchers and 3D developers with a 24GB+ NVIDIA GPU
Strengths
- Handles open surfaces, non-manifold geometry and internal structures without iso-surface conversion
- Reported H100 times: about 3s at 512³, 17s at 1024³, 60s at 1536³
- Full training code and data toolkit included, plus shape-conditioned PBR texturing
- MIT license for code and model; weights on Hugging Face with a hosted demo
Weaknesses
- Needs an NVIDIA GPU with at least 24GB memory; tested only on A100 and H100
- Tested only on Linux; setup compiles CUDA packages and may be slow
- GLB exports as OPAQUE; transparency must be wired up manually in 3D software
- nvdiffrast and nvdiffrec dependencies carry separate license terms
What it needs
- GPU required
- Needs CUDA Toolkit 12.4, PyTorch 2.6.0, Conda, flash-attn, nvdiffrast, nvdiffrec, CuMesh, FlexGEMM
- Models: TRELLIS.2-4B
Also in Open models
See all 34| Rank | Project | Score |
|---|---|---|
| 1 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 84 out of 100 |
| 3 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 75 out of 100 |
| #4 | speechPyTorch toolkit for training and deploying ASR, TTS and speech LLM models | 70 out of 100 |
| #5 | yueLyrics-to-song model that writes an editable melody and chord score first | 65 out of 100 |