Extensions/ComfyUI_SenseNova_U1
ComfyUI Extension

ComfyUI_SenseNova_U1

SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-Unify Architecture

By smthemex·Created 7 months ago·Updated 4 days ago· 92
smthemex/ComfyUI_SenseNova_U1
Nodes2
On cloudLocal install
CategorySenseNova
Stars92
Updated4 days ago
Readme

ComfyUI_SenseNova_U1

SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-Unify Architecture

Update

  • 适配高版本transformer,精简代码 / fitting to transformer 5.4.0,simplify code
  • Add U1.5 support 新增U1.5支持(GGUF/INT8/FP8),模型缓慢上传hf,获取去我的云盘拉取
  • 新增fp8和convrot int8模型支持(Infographic-V3),support fp8 and convrot int8 models
  • 暂时以同步卸载模式适配moe模型,moe模式12G显存, 文生图时,count设置为8-9,图生图设置为4左右..
  • support SenseNova-U1-A3B-MoT-SFT and SenseNova-U1-A3B-MoT and SenseNova-U1-A3B-MoT-SFT-gguf ,支持A3B MOE的单体合并模型和gguf模型,快速测试可以修改节点加载repo的分割模型.
  • If Vram >16G make prefetch_count =0, 显存大于16G时,设置swap(prefetch_count)数值为0以关闭层交换(使用Q6 gguf时)

1.Installation

In the ./ComfyUI/custom_nodes directory, run the following:

git clone https://github.com/smthemex/ComfyUI_SenseNova_U1

2.requirements

pip install -r requirements.txt

3.checkpoints

U1.5-8B-links (preview and official )
A3B-links-modelscope
8B-links
A3B-links lora
夸克网盘

├── ComfyUI/models/gguf/ # GGUF模型可选
|     ├── SenseNova-U1.5-8B-MoT-SFT-Q8_0.gguf # optional 可选U1模型 SenseNova-U1-8B-MoT-8step-Q6_K.gguf or  SenseNova-U1-A3B-MoT-SFT-Q4_K_S.gguf
├── ComfyUI/models/diffusion_models/
|     ├── SenseNova-U1.5-8B-MoT-SFT_fp8_scaled.safetensors # 可选U1模型
├── ComfyUI/models/loras/
|     ├── SenseNova-U1.5-8B-MoT-LoRA-8step-V2.safetensors # optional 可选U1模型  SenseNova-U1-8B-MoT-LoRA-8step-V1.0.safetensors 

  1. Example

5 .Citation

@misc{sensenova2026neounify,
  title        = {NEO-unify: Building Native Multimodal Unified Models End to End},
  author       = {SenseNova},
  journal      = {Hugging Face blog},
  url          = {https://huggingface.co/blog/sensenova/neo-unify},
  year         = {2026}
}

@article{sensenova2026sensenovau1,
  title        = {SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture},
  author       = {Diao, Haiwen and Wu, Penghao and Deng, Hanming and Wang, Jiahao and Bai, Shihao and Wu, Silei and Fan, Weichen and Ye, Wenjie and Tong, Wenwen and Fan, Xiangyu and others},
  journal      = {arXiv preprint arXiv:2605.12500},
  year         = {2026}
}