v2update: Small workflow changes, better performance, faster generation time, updated ip_adapter nodes."

The workflow can generate an image with two people and swap the faces of both individuals.

Firstly, a mask is automatically generated which controls the respective deviation IP adapter.

An initial image is generated, the faces are automatically recognized, and a face swap is performed with Reactor. Then, the faces are enhanced with two face detailers, each controlled by their own face_id models, and finally, the finished image is upscaled.

Known bugs include it not working if there are more than two people in the picture.

Also, if the individuals are standing too close together, it might result in the wrong faces being swapped.


If you're having trouble with the installation of the ip_adapter, check out the GitHub page and read the installation instructions and the troubleshooting articles.
https://github.com/cubiq/ComfyUI_IPAdapter_plus