ComfyUI_CrossImageAttention
★ 4
ComfyUI跨图像注意力参考图像风格迁移
将CrossImageAttention集成到ComfyUI,在生成流程中对参考图与目标图应用跨图像注意力,实现参考式风格或细节迁移与精确引导。
💡 在ComfyUI中用参考图对生成进行跨图像注意力引导
🍴 1 Forks💻 Python🔄 2024-08-16
https://pan.quark.cn/s/e58c8376a81b
📦 requirements.txt
matplotlib==3.6.3
matplotlib-inline==0.1.6
jupyter==1.0.0
numpy==1.24.1
pyrallis==0.3.1
torch==2.0.1
torchvision==0.15.2
diffusers==0.19.3
transformers==4.30.2
accelerate==0.20.3
huggingface-hub==0.16.4
xformers==0.0.21
tokenizers==0.13.3
nltk==3.8.1
Pillow==10.1.0
scikit_learn==1.3.0
tqdm==4.64.1
📄 README
ComfyUI nodes to use CrossImageAttention
https://github.com/garibida/cross-image-attention
Install
推荐使用管理器 ComfyUI Manager 安装(On the Way)
手动安装:
cd custom_nodes
git clone https://github.com/leeguandong/ComfyUI_CrossImageAttention.git
重启 ComfyUI
Use
节点:
Load Image Path: 加载图片路径到config
CIA Config: 配置config文件
Appearance Transfer Model Loader: 模型加载
Load Latents: 加载图片的latent,提供后续的QKV
Appearance Transfer Inference: 模型推理
Stars
[](https://star-history.com/#leeguandong/ComfyUI_CrossImageAttention&Date)