This repository is a ComfyUI port of the WithAnyone model introduced in the paper

WithAnyone: Towards Controllable and ID-Consistent Image Generation (2025)

Original implementation: Doby-Xu/WithAnyone.
Huge thanks and congratulations to the authors for their excellent work and for releasing it as open source.

Clone this repository under your ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/okdalto/ComfyUI-WithAnyone
cd ComfyUI-WithAnyone
pip install -r requirements.txt

Go to ComfyUI/models/clip and download the following:

wget https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
wget https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors

Go to ComfyUI/models/diffusion_models and download:

wget https://huggingface.co/bstungnguyen/Flux/resolve/main/flux1-dev.safetensors
wget https://huggingface.co/WithAnyone/WithAnyone/resolve/main/withanyone.safetensors

Go to ComfyUI/models/vae and download:

wget https://huggingface.co/bstungnguyen/Flux/resolve/main/vae/diffusion_pytorch_model.safetensors

Go to ComfyUI/models/diffusers and download the multilingual SigLIP model:

apt update
apt install git-lfs -y
git lfs install
git clone https://huggingface.co/google/siglip-base-patch16-256-multilingual

The WithAnyone Model Loader will automatically download the ArcFace model into
ComfyUI/custom_nodes/ComfyUI-WithAnyone/models/.

However, there is a known issue. If you encounter an error such as:

assert 'detection' in self.models

please manually move the downloaded model directory to the correct location under the path above.

1. Load workflow by the given workflow.png or workflow_demo.json files.
2. Upload 1~4 reference images, each containing a single face.
3. [Recommended] Use bbox calculation nodes to determine where to place the faces in the output image.
4. [Recommended] Adjust siglip_weight in the Sampler node. Higher values yield better identity consistency but may reduce adherence to the text prompt.
5. [Recommended] Add a LoRA like in workflow_lora.png