ComfyUI custom nodes for TRELLIS.2 - Microsoft's state-of-the-art image-to-3D generation model.
Generate high-quality 3D meshes with PBR (Physically Based Rendering) materials from a single image.
Search for "ComfyUI-TRELLIS2" in ComfyUI Manager and click install.
trello_30fps.mp4
- Python 3.10+
- PyTorch 2.0+
- CUDA-compatible GPU with 8GB VRAM (16GB+ recommended)
All dependencies should theoretically install through wheel by running install.py! This node was made for comfyui-manager install, but you can also pip install requirements and "python install.py" manually.
The video render nodes will be implemented soon! Not yet working.
Models are downloaded automatically from HuggingFace on first run. If you need to download manually:
From microsoft/TRELLIS-image-large/ckpts:
ss_dec_conv3d_16l8_fp16.json+ss_dec_conv3d_16l8_fp16.safetensors
From microsoft/TRELLIS.2-4B/ckpts:
ss_flow_img_dit_1_3B_64_bf16.json+.safetensorsshape_dec_next_dc_f16c32_fp16.json+.safetensorsslat_flow_img2shape_dit_1_3B_512_bf16.json+.safetensorsslat_flow_img2shape_dit_1_3B_1024_bf16.json+.safetensorstex_dec_next_dc_f16c32_fp16.json+.safetensorsslat_flow_imgshape2tex_dit_1_3B_512_bf16.json+.safetensorsslat_flow_imgshape2tex_dit_1_3B_1024_bf16.json+.safetensors
Place files in: ComfyUI/models/trellis2/ckpts/
ComfyUI/models/trellis2/
└── ckpts/
├── ss_dec_conv3d_16l8_fp16.json
├── ss_dec_conv3d_16l8_fp16.safetensors
├── ss_flow_img_dit_1_3B_64_bf16.json
├── ss_flow_img_dit_1_3B_64_bf16.safetensors
├── shape_dec_next_dc_f16c32_fp16.json
├── shape_dec_next_dc_f16c32_fp16.safetensors
├── ... (etc)
Other HuggingFace models (auto-downloaded):
- DinoV3:
facebook/dinov3-vitl16-pretrain-lvd1689m - BiRefNet:
briaai/RMBG-2.0
Questions or feature requests? Open a Discussion on GitHub.
Join the Comfy3D Discord for help, updates, and chat about 3D workflows in ComfyUI.
- TRELLIS.2 by Microsoft Research
Contributions are welcome! Please feel free to submit issues and pull requests.
Thank you @fire for pointing out the segment_reduce bug!